Chapter 7
7. SURVEY OF SHINGLE REPLACEMENTS
The frequencies of shingle replacements in the hail areas depicted in Fig. 12 were investigated. Asking an insurance company for a list of clients who submitted claims for roof damage was considered, but no insurance company was likely to provide this confidential information; consequently, the records of 2 local roofing companies were perused.
a. Motivation for studying reshinglings
Insurance and reinsurance firms are greatly interested in the frequency of severe, and thus costly, urban hailstorms (Armstrong, 1993). For instance, BEP International in Toronto and Sedgwick Payne Insurance Strategy, Inc. in Seattle are involved with in-house hailstorm research programs. The Alberta Severe Weather Management Society, formed by a group of insurance firms operating in Alberta, was created in response to the loss caused by the 1991 Calgary hailstorm. In 1996, the group commenced a 5 year program of cloud seeding near Calgary and Red Deer (Rogers, 1996; Renick and Rogers, 1996). Furthermore, the Insurance Institute for Property Loss Reduction (IIPLR) in Boston commissioned Underwriters Laboratories Inc. to study the resistance of shingles to impacts (Insurance Institute for Property Loss Reduction, 1995; Laymon and Rhodes, 1995). In the spring of 1996 at Bismark, North Dakota, representatives of the insurance industry met with cloud physicists and hail suppression experts to review the state of hail science (Boe, 1996).
The total insurance settlement for the July 31, 1987 storms was $250 million, then a record amount for a Canadian natural disaster. Figures deduced by Charlton et al. (1995) suggested that approximately $67 million was paid to 32000 homeowners, principally for hail damage to shingles. Commercial policy holders collected $150 million, mostly for wind damage along the tornado path, and most of the remaining $33 million was paid for hail damage to automobiles. The largest insurance loss in Canada, $400 million, was caused by the September 7, 1991 hailstorm in Calgary (Charlton et al., 1995), where $210 million was paid for hail damage to houses, and most of the remaining $190 million was paid for hail damage to automobiles. More detailed accounts of the insurance losses caused by this storm were given in Charlton et al. (1995) and Charlton and Kachman (1996). Estimates provided by Alan Wood suggested that the insured loss caused by the floods in the Saguenay region of Quebec in July, 1996 may exceed the record set by the Calgary hailstorm.
During the summer of 1996, destructive hailstorms struck Calgary twice and Winnipeg once. Alan Wood estimated that the insurance losses in Calgary on July 16 and July 24 totalled $160 million, though the damage was apparently reduced by the cloud seeding program (Canadian Press, 1996; MacLean, 1996). Maps prepared by Dudley (1996) of the Calgary weather office showed that the 2 damage swaths overlapped in south Calgary. On July 16, a hailstorm buffeted Winnipeg and caused $100 million of insured damage (Chalmers, 1996). As Winnipeg was struck one-half hour before Calgary, it is not possible that the two cities were struck by a typical weather system moving from west to east.
A detailed account of the hail suppression operations in the Calgary-Red Deer region during 1996 was prepared by Weather Modification Inc. (Krauss, 1996) for the Alberta Severe Weather Management Society. It indicated that the project was highly successful from an operational standpoint, Calgary hailstorms notwithstanding.
b. Determining the frequency of urban hailstorms
Cities are too small and widely separated to be frequently buffeted by large quantities of giant hail; therefore, a satisfactory climatology cannot be developed from such rare incidents. However, it can fairly safely be stated that the average interval between severe hailstorms is roughly one decade for each of the main cities in Alberta.
Throughout the world, agricultural regions where severe hailstorms are common usually have excellent records of crop insurance claims; these records cover many decades but, typically, do not include notes about the observed hail size or exact swath dimensions. For example, Willemse (1995) studied many years of records of crop damage for a region of Switzerland but detailed tabulations of hailfall parameters did not exist. Furthermore, crop damage is, at best, weakly related to the fall of large hail (Summers and Wojtiw, 1971). Vast quantities of pea- and grape-sized hail, in combination with strong wind, can devastate many types of crops, but hail of these sizes is unlikely to inflict substantial damage upon houses or automobiles.
A crude relationship between reported hailstone sizes and shingle damage can be established from the field observations recorded in the present study or by laboratory experiments involving impacts to both new and weathered shingles. If available, local historical records of reported hail sizes and hail swath dimensions might then be employed to estimate the extent and frequency of future severe, urban hailstorms. Weather radar records may also be used to augment ground-based observations of hail (Al-Jumily et al., 1991; Balakrishman and Zrnic, 1990; Holt et al., 1994; Schiesser et al., 1995; Wrenshall, 1978), but the relationships between radar echo parameters and the sizes and quantities of hailstones need further investigation.
Some regional hail databases were developed by soliciting hail reports from farmers. These databases, augmented by studies of crop insurance, were usually assembled to evaluate the effectiveness of hail suppression programs like that of the Alberta Hail Project (Alberta Research Council, 1968-1985, 1986). For example, the Alberta Research Council collected more than 85000 hail reports from farmers located between Calgary and Edmonton in central Alberta (Wojtiw, 1987). These farmers' reports indicated that when hail had occurred, the frequency of maximum size was pea 44%, grape 37%, walnut 13%, golfball 5%, and larger than golfball 1.5% of the time. The reports are archived at the Department of Earth and Atmospheric Sciences at the University of Alberta (Kochtubajda et al., 1996). Studies of the frequencies of hailstorms which produced large and giant hailstones in the United States were conducted by Kelly et al. (1985) and Sammler (1993) but similar studies have not been conducted for Canada.
c. Maps of shingle replacements
In the spring of 1994, the invoice records of the A Clark Shingle Company were examined for 1987, 1988, and 1990, with the permission of Doug Clark. Each invoice included the type of work performed and the address of the client. These records distinguished between minor repairs, installations of shingles on new houses, and replacements of shingles on houses, referred to as T & C's (Tear off the old shingles, install new shingles, and Clean the debris) on the invoices.
The dots on Fig. 16 mark the locations of 332 T & C's completed between August 1987 and July 1988. These 332 T & C's were a significant proportion of the total number of T & C's performed by the shingle company in that period. All T & C's completed in this period were counted and examined, but to maintain business confidentiality, the total will not be given. Furthermore, examination of the invoices and discussions with management at the A Clark Shingle Company were convincing: the company services roofs throughout Greater Edmonton and does not dominate in or exclude itself from some areas. Thus, the pattern of T & C's shown in Fig. 16 should be a reasonable facsimile of the spatial distribution of all reshinglings performed by roofing companies in Greater Edmonton. Also displayed in Fig. 16 are the maximum hail-size boundaries from Fig. 12 and the residential areas (shading) from Fig. 2.
Figure 17 shows the locations of 125 T & C's completed by the A Clark Shingle Company in 1990. The ALWC did not record an instance of large hail falling in Greater Edmonton in 1988, 1989, or 1990 (Alberta Weather Centre, 1984-1996). Figure 17 was included so that the distribution of reshinglings in 1990 could be compared with the distribution for the 12 months following the tornado ('87-'88). The fraction of 1990 T & C's plotted in Fig. 17 is the same as the fraction plotted in Fig. 16. Thus, the number of T & C's performed by the company in the 12 months following the tornado was 2.7 times larger than the number completed in 1990. The monthly average of reshinglings for the 7 months prior to the tornado and the 5 and 12 months following the tornado were also calculated. The ratios of the 5 and 12 post-tornado monthly averages to the average of the 7 pre-tornado months were 4.2 and 3.4, respectively: an impressive increase in business activity!
The increase in the number of T & C's that the shingle company could perform in the 12 months after the tornado might have been constrained by several factors: the availability of new roofing crews, the maximum number of hours that crews could work, the emergence of independent shingling crews in the city, and the number of new houses under construction that also required shingles. In the spring of 1995, Charlton spoke with a shingle installer in his late twenties who has spent most of his adult life working in cities which had been pummeled by severe hailstorms; the hailstorms in Greater Edmonton in 1987, Medicine Hat and Calgary in 1988, Calgary and Red Deer in 1991, Calgary in 1992, and Salmon Arm, British Columbia, on August 8, 1994, had been the man's principal source of employment. He reshingled roofs in Salmon Arm until the spring of 1995. These and other costly insurance events in Canada were listed in Facts of the General Insurance Industry in Canada (Insurance Bureau of Canada, 1996). In the spring of 1997, Charlton spoke with another shingle installer who laughed when he recalled his first hail damaged roof; the hastily assembled crew had to read the installation instructions printed on the shingle bundles!
d. Tabulation of shingle replacements by maximum hail-size categories
Table 3 displays information about reshinglings derived from Fig. 16 and Fig. 17. The first and second columns of the table give the total and residential areas, respectively, of the walnut, golfball, tennis ball, and "Beyond Large Hail" areas in Greater Edmonton. These areas were first given in Section 5. The area of tennis ball-sized hail was sub-divided into 4 residential zones. Data for these sub-regions are also provided in Table 3. Kaskitayo is the residential region immediately to the west of Mill Woods, across the Calgary Trail commercial corridor (8 E), but its area given in Table 3 includes only that part of Kaskitayo within the tennis ball boundary. The residential areas of Greater Edmonton did not change significantly between 1987 and 1990, although the number of houses (detached, duplex, and multi-family) had increased by 9000. Thus, the residential areas were appropriate for both '87-'88 and 1990.
The third and fourth columns of Table 3 show the number of T & C's within each hail-size boundary from Fig. 16 and Fig. 17, respectively. Only the areas within the tennis ball boundary had large differences in the numbers of T & C's between '87-'88 and 1990, particularly West Mill Woods and Kaskitayo. West Mill Woods was defined as the western part of Mill Woods which contained all but 6 of the T & C's plotted in that sub-division (Fig. 16).
e. Shingle replacement per unit of residential area
For each region, the number of plotted T & C's for '87-'88 was divided by the residential area, thus giving the spatial density of plotted T & C's; the results are shown in the fifth column of Table 3. For comparison, the figures for 1990 are provided in the sixth column.
For the 12 months following the storms ('87-'88), the density of T & C's (the number of plotted T & C's per square kilometre) within the walnut area, 0.6, was not substantially different from the density in the region outside the boundary of walnut-sized hail, 0.5. The density of reshinglings in the golfball area, 1.4, was more than twice that in the walnut area; in the region of tennis ball- sized hail, the density, 4.2, was 3 times larger than that in the golfball region.
In 1990, the range of T & C densities was far smaller than in the year following the tornado. The largest densities were in the regions of walnut (1.0), golfball (1.0), and the area of tennis ball outside of Mill Woods and Kaskitayo (1.2). This suggested that many homeowners not residing in Kaskitayo or West Mill Woods were beginning to find shingle damage caused by the storms of July 31, 1987, or that they had decided to spend their insurance settlements collected after the storms. In the 12 months following the storms, West Mill Woods and Kaskitayo had, by far, the highest rates of reshingling per unit residential area, but their rates were among the lowest in 1990. Apparently, repairs to roofs in these 2 regions were needed almost immediately.
f. Estimating the reshingling rates
The seventh column of Table 3 shows the estimated percentages of houses that were reshingled in each of the hail areas in '87-'88, that is, the real reshingling rates. To determine these rates, 4 assumptions were made.
1) All of the 32000 householder insurance claimants in Greater Edmonton had their houses reshingled in the 12 months following the tornado.
2) The 32000 reshinglings done for the insurance claimants were the only reshinglings performed in Greater Edmonton from August 1987 through July 1988.
3) For each area listed in Table 3, the number of houses reshingled was proportional to the number of T & C's plotted in it.
4) The spatial densities of houses within postal areas were constant.
The first assumption led to an exaggeration of the percentage of reshinglings in each region. The second assumption led to under- estimates, but the number of routine reshinglings performed in '87-'88 was probably much lower than if the storms had not occurred. The third assumption was based upon the belief, defended earlier, that A Clark Shingle Company's market share did not vary appreciably across the city.
A detailed explanation of the fourth assumption is warranted. Canada Post divided Greater Edmonton into 40 Forward Sortation Areas (FSAs) and tabulated the number of houses, farms, apartment units, and businesses in each. These data were used to estimate the number of houses in each hail area or part thereof. For most FSA's the average number of houses per unit of residential area was between 800 and 1400 per km2. The average for Greater Edmonton was 976 homes per km2. Surprisingly, the differences among the averages for the hail areas (as opposed to FSA's) were small. Thus, for '87-'88, reshingling rates in percent (column 7) were proportional, by a factor of approximately 10, to the numbers of T & C's per km2 (column five). For 1990, reshingling rates (column nine) to one significant digit obscured the nearly constant factor of 5.7. These factors implied that the numbers of houses per unit residential area within each hail area were nearly constant, despite the variations among FSA's discussed previously. Thus, the comparisons of the densities of T & C's, discussed in subsection e, also applied to the reshingling rates in percentage but, before moving on to the next section, it must be mentioned that reshingling rates of 77% in Kaskitayo and 89% in West Mill Woods are certainly remarkable.
g. Confirmation of the distribution of T & C's in '87-'88
The eighth column of Table 3, labelled Alf's, gives a second estimate of the percentage of reshinglings in each area in '87-'88. In 1995, when the analysis of T & C data from A Clark Shingle Company was nearly complete, Alf's Roofing was contacted. Surprisingly, the owner, Alfred Weimann, had retained the on-site estimates that his employees had completed in 1987. Unlike an invoice, an estimate usually included a description of the damage to the roof, the general condition of the roof, and the cause of the damage. An estimate provided other information as well: whether a repair or reshingling was recommended, the presence of structural damage, whether Alf's Roofing received the contract, and, occasionally, the comments of the homeowner. The locations where Alf's Roofing recommended a reshingling were used to supplement the information from A Clark Shingle Co. Only a modest fraction of the estimates was examined, and the locations of 319 "recommended reshinglings", were plotted on a map. Also recorded were the locations where only minor repairs were needed, where old and worn shingles were found, where hail caused structural damage to roofs, or where the estimator found no evidence of hail or wind damage. Alf's "recommended reshinglings" were used to determine the percentages of houses needing reshingling. Again, the 4 assumptions discussed in subsection f were used to determine the percentages.
Concern about the validity of using the distribution derived from A Clark Shingle Company's invoices to represent all reshinglings in Greater Edmonton was allayed by the strong similarities among the real reshingling rates and the recommended reshingling rates (columns seven and eight, respectively, of Table 3). Proportionately, the only large difference was for East Mill Woods, 7% from Clark's and 15% from Alf's. Possible reasons for this will be discussed in later subsections.
An examination of the damage reports (Fig. 15) from respondents to the newspaper survey was completed. In the tennis ball area the percentage reporting hail damage to roofs (labelled 'r' or 'h') was calculated for each sub-region: for Kaskitayo, 40%; for West Mill Woods, 44%; for East Mill Woods, 19%; and for "Other tennis ball area", 16%. Like the reshingling values derived from the records of the roofing firms these values also suggest that reshinglings caused by hail damage were common in Kaskitayo and West Mill Woods but decidedly less frequent in East Mill Woods and the "Other tennis ball area". But reports of wind damage to roofs (labelled 'R' or 'H') were common from participants in East Mill Woods (28%). The respondents in East Mill Woods who lived near the path of the tornado might have been more inclined to examine their roofs and complete a newspaper survey than participants in the other 3 tennis ball areas. For the entire tennis ball area, the percentage of reports of damaged roofs (wind and hail) was 36%, a value similar to the percentages from A Clark Shingle Co. (41%) and Alf's Roofing (40%).
h. Reshinglings in 1990, a quiet hail year
The ninth column of Table 3 shows the estimated percentages of houses reshingled in Greater Edmonton in 1990. These percentages were calculated in the same way as those given for '87-'88 but there was one value which had to be estimated, namely, the percentage of homes reshingled in Greater Edmonton during 1990. Three methods were used to estimate the evasive value appropriate to 1990.
Assuming that A Clark Shingle Co. had the same market share in 1990 as in '87-88 suggested that 6.3% of homes were reshingled in 1990. Assuming that shingles were replaced every 20 years gave a value of 2.3% when statistics on the age of existing houses were consulted. When production figures from a major local shingle manufacturer, BPCO, were combined with regional sales information and an estimate that 70% of production is used for reshinglings, a value of 4.4% for a fraction of Edmonton homes reshingled in 1990 was obtained. Averaging the three estimates gave the city-wide value of 4% given in Table 3.
Estimated reshingling rates for 1990, calculated for each area struck by hail in 1987, varied between 0% and 6% (column 9 of Table 3). Preliminary comparisons of these values are consistent with comments given in subsection e using the corresponding T & C densities.
i. Comparison of reshingling rates in Edmonton, 1987 with Calgary, 1991
As noted in section 5, subsection a, the estimated percentages of homeowners with insurance claims for hail damage were 18% for Edmonton and 28% for Calgary. In Calgary, one insurer, The Co-operators Insurance/Financial Services, settled claims with 34% of their policy holders (Charlton et al., 1995). Using Co- operators claims rates, derived from confidential, detailed data arranged by FSA, the total number of claims in Calgary would have been 70000. This is 20% larger than 58367, the actual number of claims according to Alan Wood of the Insurance Bureau of Canada.
In 4 of the 30 forward sortation areas (FSAs) in Calgary, Co-operators had claims rates between 82% and 92%! Assuming that a claim is equivalent to a reshingling, the values from these 4 FSAs were similar to the percentages of reshinglings in West Mill Woods (89%) and the area of Kaskitayo buffeted by tennis-ball-sized hail (77%). Charlton et al. (1995) found that the damage swath in Calgary, defined as areas designated as "many homes damaged" and "most homes damaged" by 2 Wawanesa Mutual Insurance Co. adjustors, had an areal extent of 130 km2. This swath corresponded well to the area of large hail, that is, walnut size or larger. The area of large hail in Greater Edmonton, 270 km2 (Table 3), was more than twice as large. The estimated number of houses exposed to large hail was 50% larger in Greater Edmonton - 120000 houses versus 80000 in Calgary. Furthermore, of the 638 participants from Greater Edmonton who recorded a maximum hail size, 73% reported hail of golfball size or larger, and 54% reported tennis ball or larger; but, in Calgary, only 58% of the 60 respondents to an informal survey reported golfball or larger, and 9% reported tennis ball or larger. Clearly, the area buffeted by giant hail and the number of houses in it were decidedly smaller in Calgary than in Edmonton. Was the record insurance loss in Calgary inflated by an increased willingness to file a claim?
Six FSAs in Calgary lay entirely within the swath, and portions of 7 other FSAs were within it. Using information about claims from Co-operators and housing figures from Canada Post, the claims rate for Co-operators for these 13 FSAs was estimated to be 60%, assuming that damage was evenly distributed throughout the 13 affected FSAs, or 80%, assuming that damage was limited to the damage swath determined by the adjustors. Either of these values, if interpreted as the reshingling rate in the region of Calgary struck by large hail, greatly exceeded the 24% reshingling rate for the area of large hail in Greater Edmonton (Table 3). Apparently, the willingness of homeowners in Alberta to submit a claim for wind and hail damage had greatly increased between 1987 and 1991!
An exact comparison of the consequences of the Edmonton and Calgary hailstorms was difficult because insurance claims were not organized by FSA in 1987, and reshingling information from a roofing company in Calgary was not collected. Furthermore, the number of survey respondents from Calgary (60) is just 8% of the number from Greater Edmonton (755) and, thus, insufficient for delineating hail-size boundaries.
j. The anomalous reshingling rate in East Mill Woods
The reshingling rate in East Mill Woods, which lay entirely within the tennis ball boundary (Fig. 12), was a surprisingly low 7%. The recommended reshingling rate, 15%, was also curiously small. Meteorologists would be tempted to claim that "soft" hail, the kind that splatters when it strikes lawns and, thus, could not damage shingles, fell there. The survey forms from participants in this region were carefully read, but no comment about soft or splattered hailstones was found. The maps of hailstone measurements, Fig. 13 and Fig. 14, both suggest that the hailstones in East Mill Woods were, in general, smaller than those in West Mill Woods but of similar sizes to those in Kaskitayo, the residential community west of Mill Woods. But the reshingling rate in Kaskitayo, 77%, was 11 times greater than that in East Mill Woods!
The reports of the largest hail sizes were tabulated for each of the 4 areas within the tennis ball boundary. West Mill Woods, with 61% of the hail reports indicating larger-than-tennis- ball, apparently received the worst of the hailstorm, but its reshingling rates, real and recommended, were similar to those for Kaskitayo where only 27% of respondents reported larger-than tennis ball. Surprisingly, East Mill Woods, with just one eleventh the reshingling rate of Kaskitayo, had a modestly larger proportion of reports of larger-than-tennis-ball (32%) than did Kaskitayo. The distribution of hail sizes in East Mill Woods was closest to that of "Other tennis ball areas", though the reshingling rates in these 2 areas were also substantially different, 7% versus 23% but Alf's recommended reshinglings were quite similar; 15% in East Mill Woods versus 22% in "Other tennis ball areas."
The anomalies in the reshingling rates among the areas within the tennis ball boundary could not be adequately explained by variations in the maximum-hail-size categories. The possibility that the anomalies were caused, at least partially, by another hailfall parameter, namely the most common size, was also investigated. (The Hail Report (Fig. 1b) requested that respondents record the most common sizes.) The evidence was limited, but the average of the most common sizes for a region in the tennis ball area seemed to be a marginally better indicator of its reshingling rate than the average of the largest sizes, provided that East Mill Woods was ignored.
The Hail Report (Fig. 1b) also asked for an estimate of the duration of hailfall. Kaskitayo had an average duration of 29 minutes. For West Mill Woods, hailfall averaged 24 minutes. For East Mill Woods, the average was 12 minutes. The average of the durations from "Other tennis ball areas" was 25 minutes. Thus, for East Mill Woods, hailfall, and presumably the fall of tennis ball hail, lasted approximately one half as long as in the 3 other tennis ball areas. Within the "Other tennis ball areas" there was also a tendency toward shorter hailfalls in the east. Presumably, the probability of a roof being hit and, consequently, damaged by tennis ball hail increased as the duration of hail increases. Thus, the observations of hail duration explained, at least partly, the low reshingling rate in East Mill Woods.
k. Relating the age of housing with the resistance of shingles to hail, and roof penetration by hail
The average ages of houses in Mill Woods and Kaskitayo were determined by examining Neighbourhood Fact Sheets, published by the City of Edmonton's Planning Department. East Mill Woods is, on average, a newer community than West Mill Woods or Kaskitayo, although the difference in the average ages is modest. In 1987, north Kaskitayo was a little more than 15 years old and south Kaskitayo was 5 years old; West Mill Woods was 10 years old, and East Mill Woods was 5 years old, though a few neighbourhoods were nearly 10 years of age. All other communities struck by giant hail were more than 17 years old, with the exception of those at the northernmost tip of the golfball boundary. Perhaps East Mill Woods had a low reshingling rate, in part, because the community was relatively new; there were many reshinglings in south Kaskitayo (Fig. 16), however, even though houses there were about the same age as those in East Mill Woods.
Estimators from Alf's Roofing found roofs which showed no evidence of hail or wind damage in every region of the city, including a few in Kaskitayo and West Mill Woods. A list of the locations of 10 houses where hail penetrated the roofs was made; seven locations were found in estimates from Alf's Roofing, and 3 came from survey participants. All but 1 penetration occurred in Mill Woods; four were in its extreme southwest corner. This district had the highest averages of both maximum and of most common hail sizes of any neighbourhood. The one penetration outside Mill Woods occurred at 10.5 E, 10.0 N (Fig. 15), approximately 6 km north of Mill Woods, and only 1 km southeast of the place where the second largest hailstone (264 g) was collected (Fig. 14).
l. Relating reshinglings with laboratory tests of shingles
The impact resistance of domestic roofing products has been determined by striking them with dropped steel balls (Laymon and Rhodes, 1995) and ice spheres (Greenfeld, 1969; Koontz, 1991) propelled by compressed air. These engineering studies contain only limited information about the effects of weathering on the resiliance of asphalt shingles. Nevertheless, they do suggest that new asphalt shingles are fractured by 5.1 cm (2 inch) spherical hailstones, and after 10 years of weathering, 3.5 cm (1.25 to 1.50 inch) ice spheres will cause fracturing. At 15 years, 1.9 cm (0.75 inch) will suffice. At these minimum sizes, the loss of granules on the upper surface is often minimal and the fracturing of the shingle mat is often visible only on the lower surface. For equal kinetic energy of impact these three sizes were adjusted to account for the average shape of large and giant hailstones (Barge and Isaac, 1973; Charlton et. al., 1989). It was concluded that during the first five years, asphalt shingles of average quality could withstand a fall of real hailstones with maximum dimensions up to 6.4 cm (actual tennis balls). After 10 years of weathering, stones with maximum dimensions of 4.4 cm (actual golfballs) would fracture the shingles, and beyond 15 years the blows of hailstones as small as 2.3 cm (small walnuts) could cause fracturing which, in time, would lead to leaking of the roof.
Both Randy Clark of A Clark Shingle Co. and Alf Weimann of Alf's Roofing noted that weathering is the principal factor determining the susceptibility of asphalt shingles to fracturing and granular loss. Dutt (1987) studied the adhesion of granules on both new and old asphalt shingles. Samples as old as 29 years were collected from roofs in Ottawa. He found that the exposed surfaces typically lost granules at a rate of 1% per year. Dutt did not test the resiliency of the shingles to impacts.
There are additional, secondary factors affecting the resilience of asphalt shingles to impacts: thickness, manufacturing techniques, variations among batches, and, no doubt, the pitch and orientation of the roof; but, if weathering is the key consideration, the tests conducted by Underwriters Laboratories should have been, at the very least, extended to include testing shingles which have been weathered for at least 10 years.
The retention of granular coatings and hidden fracturing might have been the reasons why the reshingling rates in 1990 (Table 3) were high in the walnut and golfball areas, relative to the area "Beyond Large Hail". Perhaps many homeowners inspected their roofs soon after the storms and failed to find any evidence of damage, but the mats of their shingles had cracked, and, within 3 years, their roofs began to leak. Greenfeld (1969) suggested that damage to shingles could be undetectable shortly after a hailstorm but that the fall of modest size hail months or years later or ice penetration during winter could make the damage noticeable.
In 1990, 18 of the 125 plotted T & C's (Fig. 17) were in communities that were less than 20 years old. Of these 18 seemingly premature reshinglings, 13 were in communities buffeted by large hail in 1987. Conversely, Alf Weimann noted that he had seen roofs with shingles that had suffered dents and granular losses in 1987 but remained functional in 1995. Many insured homeowners, however, will not accept the risk of developing leaks in the future, and, because policy holders have 1 year to proffer claims to their insurance companies, they will submit their claims, pressure their insurance companies for their settlements, and replace their shingles as soon as possible.
m. Summary of shingle replacements
Kaskitayo and West Mill Woods were buffeted by tennis-ball-sized hail, and the reshingling rates in these 2 areas, derived from the examination of the invoices of A Clark Shingle Co., were estimated to be 77% and 89%, respectively. Using estimates from Alf's Roofing, the rates of recommended reshinglings in these 2 areas were calculated to be 84% and 80%, respectively. The houses in both communities were, on average, 10 years old in 1987. The roofs of several houses in West Mill Woods were pierced, but there were apparently no roof penetrations in Kaskitayo. The distributions of reports of giant hailstones (golfball or larger) from East Mill Woods and Kaskitayo were quite similar. But in the newer community, East Mill Woods, the percentages of houses that were reshingled, 7%, and needed reshingling, 15%, were far smaller than in Kaskitayo. This discrepancy was explained, at least partially, by the averages of the reported hail durations in these 2 communities: 29 minutes in Kaskitayo and 12 minutes in East Mill Woods. Furthermore, the relatively new houses in East Mill Woods were probably less susceptible to shingle damage by tennis-ball- sized hail than the houses in older communities, a possibility supported by laboratory experiments which implied that new asphalt shingles (produced in this decade) could withstand a blow from a small tennis- ball-sized hailstone. Comparison of the laboratory experiments also suggested that the resistance of commonly-used asphalt shingles to impacts had increased between 1969 and 1995. Whether soft hail fell in East Mill Woods could not be confirmed.
One variable has not been investigated to date; namely, the increasing popularity of oriented strand board (OSB) as a roof sheathing material. It is reported to be much harder and stronger than plywood. Dent-resistant OSB is believed to have been used extensively on the newer roofs in East Mill Woods. One source at A Clark Shingle Co. believed that all hail-penetrated roofs were sheathed in plywood but an acquaintance living in central Mill Woods saw unshingled OSB penetrated on a nearby home under construction. After 10 years, the acquaintance has not replaced his dented shingles.
For 4 FSAs in Calgary in 1991, The Co-operators Insurance/Financial Services had annual claims rates for wind and hail damage which exceeded 80% (Charlton et al., 1995). Assuming that every claimant had his house reshingled, the reshingling rates in these areas were similar to the reshingling rates in West Mill Woods and Kaskitayo. However, the hailstones that fell in Calgary on September 7, 1991 were, on average, decidedly smaller than the ones that fell in Edmonton on July 31, 1987.