Chapter 4
4. SIGHTINGS AND WARNINGS
The Tornado Report in the survey form (Fig. 1b) includes sketches of 5 funnel types: rope, cone, columnar, smoke-like, and multiple. Figure 6 shows the funnel types observed by the respondents as the digits '1' through '5'. Generally, the larger the digit, the more severe the funnel type. The positions of the digits in Fig. 6 accurately depict the positions of the participants when they saw the tornado. After examining the distribution of funnel types shown in Fig. 6, one is likely to pose 2 questions. Firstly, what did observers in west Edmonton really see? Secondly, what tornadic phenomena were reported as multiple funnels, that is, code 5?
a. Observations of funnel clouds other than the Edmonton tornado
All reports of funnel clouds from participants west of 11.0 E were examined in detail. For most of these reports, the times of the observations and the directions from the respondents to the phenomena were determined. These directions and times are placed beside the plotted shape categories shown in Fig. 6. Most digits without a time or direction arrow are observations consistent with the passing of the main tornado. The city is quite level, and, not surprisingly, people could see the tornado from a distance of several kilometres. Several participants saw the main tornado from positions between 8.0 E and 10.0 E, an average distance of 5 km (3 miles) from the path. One report from the western edge of the city (1.9 E, 9.1 N) was approximately 16 km from the tornado. This was the most distant observation which could have been a sighting of the main tornado. The ALWC maintained a tornado warning until 1900 MDT, nearly 3 hours after the tornado dissipated. The caution of the ALWC was justified. The last funnel cloud observation by a survey participant was made in north Edmonton (10.1 E, 15.0 N) at 1830 MDT. It is likely that citizens were paying great attention to the sky after the tornado had dissipated because accounts of damage and death were being broadcast, and the tornado warning remained in effect. Most of the funnel cloud reports from west Edmonton were clustered around 2 periods: 1600 MDT, roughly the time when the tornado dispersed, and 1800 MDT, about the time when a second storm buffeted west and central Edmonton with hail, rain, and strong winds. None of these observers in west Edmonton, however, claimed to have seen a funnel cloud touch down, excluding the main tornado. Surprisingly, none of the respondents reported seeing the spin-off tornadoes depicted in Fig. 5, though the Stannards noted the damage done to their barn in their returned survey (Charlton et al., 1989).
The tornado list from the 1987 Summer Severe Weather Program (Alberta Weather Centre, 1984-1995) included 5 entries for July 31: the Edmonton tornado, rated as an F4 on the Fujita Tornado Intensity Scale (Fujita, 1973), a tornado ranked as an F2 that moved through the countryside from south of the city to east of it, and three F0 tornadoes. These F0 tornadoes were spotted to the north, northwest, and southwest of Edmonton, but none was close enough to be seen from the city. Curiously, the only reference to funnel cloud sightings in Greater Edmonton, other than those for the main tornado, is in a single sentence, found on page 41, in the Hage Report (Hage, 1987a). The sightings of funnel clouds other than the main tornado, displayed in Fig. 6, are evidence of the severity of the storms in Greater Edmonton during the few hours following the main tornado.
b. Observations of the Edmonton tornado
All but 6 of the sightings of multiple funnels, digit '5', are enclosed by 2 lines in Fig. 6, and within these 2 regions the observations of multiple types outnumber any other funnel type. Respondents might have recorded seeing multiple funnels when they witnessed 1 or more of a few possible phenomena. For example, they might have seen 2 or more tornadoes, collections of rotating debris and dust, or, most likely for July 31, cloud-like vertical streaks known as suction vortices which, sometimes, circulated around the tornado. Suction vortices were clearly recorded on 2 amateur videos which were incorporated into a widely-distributed video production by the ALWC titled Edmonton Tornado - July 31, 1987 (Alberta Weather Centre, 1988). The suction vortices were often visible when the tornado was not obscured by rain, debris, or by dense cloud in the funnel. The ground was still damp from the frequent showers in the preceding days; this prevented the wind from raising dust and obscuring the tornado's cloudy features.
Photographs and video tapes were reviewed, and the funnel types that were recorded in these media were compared with the types noted in Atchison (1988), Bullas and Wallace (1987), and Wallace (1987). The photographs and tapes showed that the tornado began as a rope (digit '1') but occasionally transformed into a narrow cone (digit '2') as it moved towards Mill Woods. The southernmost codes plotted in Fig. 6 agree with these observations. Just before the tornado struck southeast Mill Woods, it was shaped like a cone, and it had numerous suction vortices moving around it. Victor Chung (1988), a Ph. D. candidate at the Department of Earth and Atmospheric Sciences, University of Alberta, used a video recording taken from east-central Mill Woods to estimate the tangential speeds of 2 co-occurring suction vortices; he concluded that these were moving at approximately 68 km/h and 94 km/h. When the tornado battered southeast Mill Woods, it was a massive, well defined but usually featureless rotating column (code '3'), though it might have been described as a broad cone as well. Figure 7 and Figure 8 are 2 previously unpublished photographs taken in southeast Mill Woods (11.9 E, 3.7 N). Figure 7 shows part of a roof being torn from a house, and Fig. 8, taken moments later, shows a massive column looming over homes near the damaged house. A few moments later the tornado lifted, as depicted by dashed lines in Fig. 6, and moved towards and above an isolated housing development (Larkspur Lake) extending eastward from northeastern Mill Woods. In this region (12.5 E, 5-6 N), spectacular suction vortices were seen revolving about the funnel cloud. Many participants who were in Mill Woods during the tornado's passage claimed that the tornado had more than 1 shape, but only the digits corresponding to the most severe of these types were plotted in Fig. 6.
At approximately 12.5 E and 6.5 N the funnel cloud touched down again, and the tornado, in the shape of a massive cone, began ravaging the industrial area with winds of at least 331 km/h, the minimum speed for the F4 category of the Fujita scale; the wind speed was estimated from the types of damage seen in the area and reported by Wallace (1987). As it moved into the north end of the industrial area (12.5 E, 10 N), the tornado changed to a column. Nonetheless, the shapes reported by respondents in the industrial area were dominated by reports of multiple funnel and cone types, though none of the viewers in Sherwood Park, east of the industrial area, recorded a multiple funnel type. Apparently, the distance to Sherwood Park was too great for people to see detailed tornado structure. A video recording of the tornado travelling between the Sherwood Park Freeway (8.3 N) and Baseline Road (10.0 N), was taken from the building which houses the ALWC (11.1 E, 9.7 N). It showed a tornado which varied between a column and a broad cone shape as it flung debris over a wide area. A study of the video by Victor Chung (1988) determined that the debris was travelling at 144 to 209 km/h as the tornado crossed the Sherwood Park Freeway, a region where the damage to industrial buildings suggested F4 winds. Bluestein and Golden (1993) discussed a variety of methods for observing tornadoes and noted several reasons why wind speeds derived from photogrammetric analysis might lead to an underestimate of maximum wind speed. Lux (1990) estimated that the minimum wind speed required to cause the tumbling of the empty oil tank at the abandoned Texaco refinery (site H in Fig. 5 and Table 1) was 200 km/h.
Good photographs or video tapes of the tornado when it was immediately south of the river valley, in the river valley, in east Clareview, and approaching Evergreen were difficult to find. This was usually attributed to increasingly heavy rainfall, confirmed by, for example, viewers located at the ALWC building. Figure 3, discussed in Section 2, was 1 of a set of photographs taken from north of the river, apparently just before the heavy rain began. Figure 6 shows many reports of a smoke-like tornado (digit '4') from both east and west of the river between 12 N and Evergreen (16 N). These reports suggested that the tornado's approach was indeed obscured by rain during its latter stages. Just 2 reports of multiple funnels were received from observers north of the North Saskatchewan River. Meteorologists from the ALWC found but 1 pair of photographs of the tornado as it entered Evergreen. This two-exposure panorama reveals an enormous wall of diffuse cloud, but not a distinct tornado. Good photographs of the Edmonton tornado, taken before it crossed the river, were included in Atchison (1988) and in 2 memorial booklets: Black Friday by Davidson and Diotte (1987) of the Edmonton Sun, and Tornado '87 by The Graphic Edge (1987).
The various shapes of the Edmonton tornado, documented in previous studies, photographs, and video tapes, were also reflected in the responses to the survey. Several participants decided that none of the 5 funnel types shown on the survey was an accurate representation of their sightings; rather than choosing the closest shape categories, they drew sketches of the tornado.
c. Warning times from various surveys
Respondents to the newspaper survey were asked to denote the types of severe weather that they encountered and to estimate the amounts of warning time, in minutes, that they received. Charlton believed that analysis of their warning times should provide insight into the efficiency of the severe-weather-warning system and, thus, assist the federal review team chaired by K. D. Hage. The team's findings are commonly called the Hage Report (Hage, 1987a). Preliminary analysis given to Hage by Charlton showed that almost 50% of respondents claimed that they had 0 minutes warning of the severe weather they encountered, and 20% claimed that they had more than 15 minutes warning. After further examination, however, Charlton concluded that attitudes towards agencies empowered to protect the citizenry, particularly the ALWC, varied widely among participants. An analysis employing warning times from only those respondents who had witnessed the tornado should be more reliable: Carter et al. (1989a) found that victims of the May 31, 1985 tornadoes in Ontario had "vivid recollection and detailed knowledge of circumstances even months after the event".
Figure 9 is a map of the warning times and locations of respondents who saw a tornado or funnel cloud. These times vary greatly, even those from the same neighbourhoods. The recorded warning times probably represented 1 of 3 intervals: warning to sighting of the tornado, warning to experiencing the tornadic winds, or sighting to experiencing the tornadic winds. As noted earlier, most tornado and funnel cloud reports west of 8 E were not observations of the Edmonton tornado; consequently, the warning times from these reports were not analyzed. Of the 212 participants located east of 8 E who saw the tornado, 62% had some warning time, and 25% received a warning broadcast by radio or television.
Warnings broadcast before the tornado dissipated came from 1 of the 4 television stations and an undetermined number (at least 6) of the 15 radio stations serving Edmonton (Hage, 1987a). Hage also provided the times when 14 of the 19 broadcasters first received notifications of the tornado; 13 of these 14 broadcasters learned about the tornado in the first 30 minutes of its existence. This left an ample amount of time to warn most of the city. Nonetheless, many of these 13 broadcasters were slow to report the existence of the tornado to their audiences. Hage (1987a) did not discuss the reasons for this discrepancy. Many of the newspaper survey respondents complained that the stations they were patronizing did not notify them of the tornado until it had passed their neighbourhoods; numerous participants protested that some stations maintained regular programming throughout the lifetime of the tornado.
Results from a telephone survey conducted 1.5 months after the tornado (Hage, 1987a) indicated that 80 of the 100 people surveyed in east Edmonton had been aware of the approaching tornado, and that 42% had been warned via radio or television. The question about broadcast warnings in the telephone survey, like that in the newspaper survey, did not differentiate between warnings carried by radio and warnings carried by television.
Sixteen months after the tornado, Caine (1989) interviewed 77 people from Evergreen and 48 people from east Clareview to assess their recoveries from the disaster. Of the 125 people surveyed (97% owned their own homes), 65 claimed to have been in the path of the tornado, 110 had some property damage, and 52 suffered over 25000 dollars of damage. When asked "Did you get any warning of the tornado?", just 17 of the 125 respondents replied "Yes"; therefore, 86% of the 125 respondents were not aware that the tornado was approaching their homes. The 17 people who were forewarned were asked "How were you warned?" and given a choice of answers: 7 saw the approaching tornado, 10 were warned by friends, some by telephone calls, and just 3 heard a warning carried by the radio. The 108 people who had no warning were not asked whether they had a radio or television turned on during the tornado's lifetime.
Hage (1987a) conducted an on-site survey between August 10 and September 20. He interviewed both staff and managers from 18 industries struck by the tornado. At least 9 managers believed that their employees had been forewarned, including 2 who mentioned receiving warnings by radio.
The findings of these 4 surveys are compared in Table 2. The percentages of respondents who were forewarned varied greatly, 14% to 80%. Similarly, the percentages of respondents who were forewarned by a radio or television broadcast ranged widely, 2% to 42%. These values yield no firm conclusions about warning times or the effectiveness of the electronic media in disseminating the warnings.
d. Warning times from epidemiological studies
Table 2 also gives information about warning times taken from 3 epidemiological studies of tornadic events: the Edmonton tornado (Carter et al., 1989b), the tornadoes in southern Ontario on May 31, 1985 (Carter et al., 1989a), and the tornadoes in Alabama and Georgia on March 27, 1994 (Schmidlin and King, 1995). The study by Schmidlin and King (1995) placed greater emphasis on determining perceived warning times than did the other 2 epidemiological studies, and, thus, it is likely to be of greater interest to most meteorologists. One week after the tornadoes passed through rural areas of Alabama and Georgia, Schmidlin and King drove through the affected areas. They were attempting to contact anyone who had been injured, slightly or seriously, or could describe the circumstances pertinent to someone who was no longer capable of responding because of serious injury or death. Schmidlin and King asked the same set of questions used by Carter et al. (1989a). The epidemiological studies emphasized the cause and type of injuries experienced by their respective participants. For deceased respondents, the causes of deaths were determined by the appropriate medical authorities.
The efficiency of the dissemination of weather warnings to the public varied considerably among the 3 regions discussed in the epidemiological studies. Carter et al. (1989b) believed, apparently incorrectly, that there were no warnings broadcast by cable companies or television stations before the Edmonton tornado dissipated. Listeners of a popular Edmonton radio station, CJCA, received their first tornado warning at 1513 MDT (Hage, 1987a), approximately 7 minutes before the tornado reached the southeast corner of Mill Woods. Power failures terminated the broadcasts of local radio and television stations serving the region of Ontario traversed by the tornadoes (Carter et al., 1989a; Allen, 1986). As expected, Carter et al. found that none of their participants in Ontario received a broadcast tornado warning. During a conversation, Carter suggested that telephone service in the region was also disrupted. For Alabama and Georgia, the National Weather Service issued tornado warnings for 11 of the 12 counties that tornadoes passed through; these were issued typically 10 to 20 minutes before the first deaths (Schmidlin and King, 1995). Nevertheless, Schmidlin and King (1995) stated that many of the people who lived in rural areas claimed that they did not receive a broadcast warning. Still, 29% of survivors received a broadcast tornado warning, a slightly higher value than was found for tornado viewers in east Edmonton (25%). If Schmidlin and King had included uninjured persons in their survey, the percentage of interviewees who received a broadcast warning would probably be larger, though interviewing only those people who were at least slightly injured ensured that all participants had been near the tornado.
e. Warning times and respondents' positions
A deeper analysis of the warning times information gathered from the newspaper survey is displayed in Fig. 10. The 212 tornado viewers east of 8 E (Table 2) were divided into 5 groups representative of their northwards positions. Because of the small number of respondents in far-north Edmonton, the 5 groups were not allocated an equal number of participants. Each group was subdivided into 2 categories: those with television or radio assistance, and those without. Figure 10 shows that these 5 groups, from south to north, had 43, 80, 57, 26, and 6 people, and the percentages of these who received a warning via radio or television were 19, 20, 26, 54, and 0%, respectively. Thus, with the exception of the 6 respondents from Evergreen, participants in north Edmonton were more likely to receive a warning by radio or television than those south of the river. The median warning times for both categories of each group are shown graphically in Fig. 10. Median warning times were typically 10 minutes greater for those who were warned by broadcast media than those who did not receive such a warning. Median warning times for both categories increased gradually with time until the tornado approached Clareview, where it became obscured by rain; here, the warning time for those without media assistance declined precipitously to approximately one minute.
The heavy solid line in Fig. 10 depicts the tornado's location against the elapsed time since the first touchdown. It also represents the maximum possible warning times as one moves northwards, excluding the possibility of a respondent interpreting a forecast of severe weather as a tornado warning. The funnel cloud was first sighted at 1455 MDT near Leduc by Tom Taylor. He saw it touchdown for a period of 10 seconds and then promptly telephoned the ALWC (Environment Canada, 1987). The data given in Fig. 10 show that median warning times for both categories failed to increase in proportion to the growth in maximum possible warning times; nevertheless, radio and television broadcasts appear to have provided approximately 10 additional minutes of warning, and the proportion of citizens who were warned by radio or television apparently increased as the tornado travelled northward. This analysis of intra-city warning times seems to be unique for the literature about tornadoes.