Revisiting tornado watch #211 - a spatial analysis of the May 31, 1985 tornadoes using present-day data for the state of Pennsylvania

Abstract

On the evening of 31 May 1985, a devastating and deadly tornado outbreak impacted Ohio, Pennsylvania, New York, and Ontario, Canada. A total of 41 tornadoes occurred resulting in 88 fatalities: the deadliest tornado outbreak of the 1980s. Pennsylvania experienced the greatest number of tornadoes with 21 affecting the Commonwealth, causing 65 fatalities. Despite its severity, Pennsylvania’s deadliest tornado outbreak has not been spatially analyzed in academic literature. With the outbreak occurring nearly 40 years ago, one must consider the potential impacts today in comparison to the 1985 time-period. Has the natural and built landscape changed to the extent that the population would be further impacted both directly and indirectly? To assist in answering this question, this study performs a Geographic Information System (GIS) analysis of Pennsylvania’s historical and present-day land cover, census, and building data in locations impacted by the 1985 tornadoes. Data between the two time periods is compared and analyzed to quantify changes since 1985 to estimate if the impacts would be lesser or greater in relation to the expanding bull’s eye effect. This includes estimating the number of fatalities that would occur in the present-day.

Keywords:

• Tornado outbreak
• land cover
• demographics
• GIS
• Pennsylvania
• bull’s eye effect

Introduction and background

The most common and damaging type of natural disaster affecting the Commonwealth of Pennsylvania is flooding [Pennsylvania Emergency Management Agency (PEMA) 2020] as many communities are located along waterways. This follows other incidents such as fires and winter storms, emphasizing Pennsylvania’s varied natural disaster risks. However, Pennsylvania’s association with natural disasters may not necessarily coincide with one particular disaster agent but rather a specific date. That date is May 31st. On 31 May 1889, the South Fork Dam, 14 miles north of the city of Johnstown, broke due to negligence and several days of heavy rain. Ten minutes later 20 million tons of water flooded Johnstown resulting in 2,209 fatalities. It was the greatest loss of life in a single event in the history of the United States (U.S.) at that time (McCullough 1987).

A severe weather outbreak occurred on the evening of 31 May 1985 when several thunderstorms and tornadoes impacted Ohio, Pennsylvania, New York, and Ontario, Canada. A total of 41 tornadoes occurred that evening resulting in 88 total fatalities. Twenty-one of those tornadoes occurred in Pennsylvania leading to 65 fatalities and 707 injuries in the Commonwealth (Fuller 1987). Another severe weather outbreak began on 31 May 1998 and lasted until June 2nd. This tornado/derecho outbreak produced 42 tornadoes across the states of Indiana, Kentucky, Ohio, Pennsylvania, New York, New Hampshire, Connecticut and also Ontario, Canada. Pennsylvania was most impacted by the event as 45 tornadoes occurred over a three-day period, 23 on May 31st (1 fatality) and 22 on June 2nd (2 fatalities) (McCarthy and Schaefer 1999). Last, on 31 May 2002, a macroburst with 105 mile per hour (mph) winds carved a ten-mile path of destruction near the city of Pittsburgh leaving hundreds of thousands without power. It impacted Kennywood Amusement Park and knocked the roof off a wooden pavilion, leading to one fatality (Harbaugh 2021). The 1985 Tornado Outbreak is the focus of this study. What follows is a brief summary of the events that transpired.

May 1985 tornado outbreak

The summary of the May 1985 tornado outbreak is provided by the books “Tornado Watch #211” by John G. Fuller (1987) and “The Pennsylvania Weather Book” by Ben Gelber (2002). The meteorological setting on 31 May 1985 focused on a low-pressure system moving northeast through the Great Lakes region bringing unseasonably warm and humid air with dew points in upper 60s to lower 70s Fahrenheit over Ohio and western Pennsylvania. These ambient conditions coupled with a strong cold front moving south from Canada supported thunderstorm development. Despite the risk of severe weather and the issuance of flash flood watches, the weather was surprisingly calm through the middle of the afternoon with temperatures climbing into the eighties and few clouds in the sky (Gelber 2002). However, a powerful jet stream moving at 140 mph across the Great Lakes into northeastern Ohio pulled air upward very rapidly, breaking the “cap” in the atmosphere that prevented convection of warm, moist air at the surface.

At 3:30PM EDT, minor thunderstorms began to initiate in Ontario, Canada and then “as if someone threw a switch,” storm cells began to form in a squall line that stretched along the Ohio-Pennsylvania border (Fuller 1987). The National Severe Storms Forecast Center in Kansas City, Missouri issued tornado watch number 211 for the area at 4:25PM. The first fatal tornado to strike northwestern Pennsylvania formed two miles west of the Ohio border at 4:59PM (Gelber 2002). The 334-Albion tornado was an F4 that destroyed the town of Albion, Pennsylvania, resulting in twelve fatalities and injuring 82 people (Figure 1). The 337-Atlantic tornado was another violent F4 tornado that formed at 5:17PM just west of the Ohio border and traveled 56 miles through the towns of Jamestown, Atlantic, and Cochranton, subsequently causing 16 fatalities and 125 injuries. A third F4 tornado (no. 339) formed at 5:25PM in southern Erie County near Corry, Pennsylvania and traveled 28 miles into New York, injuring 12 with no fatalities. Seven more fatalities and 30 additional injuries occurred as the fourth violent F4 tornado (no. 344) developed at 6:30PM northwest of the town of Tionesta, Pennsylvania in Venango County. The 343-Wheatland tornado was the strongest (F5) and deadliest tornado of that day. It first appeared approximately 30 miles west of Youngstown, Ohio before crossing into Pennsylvania. The tornado first devastated Newton Falls, Ohio and then proceeded toward Niles, Ohio resulting in nine fatalities. The tornado then entered Pennsylvania one mile west of Wheatland, causing seven fatalities, and finally ending its 47-mile path near the town of Mercer (Gelber 2002). This F5 tornado’s winds were estimated to be more than 260 mph and resulted in a total of 18 fatalities (8 in Pennsylvania) and 310 injuries (Grazulis 1993, 1267–1268). The 350-Moshannon State Forest tornado was an interesting F4 that touched down at 7:35PM near Penfield, Pennsylvania and traveled 69 miles through the Moshannon State Forest before dissipating at 9:00PM northeast of Lock Haven. Although there were no fatalities and only one injury, approximately 88,000 trees were destroyed along with a large metal fire tower and one cabin. This tornado physically scarred the landscape as its path is evident on satellite imagery via the National Weather Service (NWS), State College office webpage (https://www.weather.gov/ctp/moshannon). Peterson (2000) studied the damage and recovery of tree species in the Moshannon State Forest with results supporting the violent nature of this tornado. He found it caused destruction over a wide area, reducing the density and basal area of trees by 84.1 and 96.6%. Further, he noted that sprouting was infrequent with 25% of snapped trees sprouting, but only 68% of those still alive 4 years later. Overall, mortality among trees in this tornado’s path was 94.1% respectively.

Figure 1. Map of 31 May 1985 Tornadoes that impacted the state of Pennsylvania.

In the end, a total of 21 tornadoes impacted Pennsylvania’s landscape during the May 1985 tornado outbreak, resulting in 75 fatalities (65 in Pennsylvania, 10 in Ohio) and 815 injuries (707 in Pennsylvania, 108 in Ohio) (Table 1). Four tornadoes started in Ohio and ended in Pennsylvania with one other starting in Pennsylvania and ending in New York. Historically, this event set several records that still stand today. It is the deadliest Pennsylvania tornado outbreak and the only time an (E)F5 tornado occurred in the Commonwealth. Pennsylvania experienced the most tornadoes ever in one day and the 65 fatalities nearly equaled the total number of fatalities (69) recorded from 1916 to 1985 [National Oceanic and Atmospheric Administration (NOAA) 1985]. The Wheatland tornado is still to date the furthest east a (E)F5 tornado has been surveyed in North America since the NWS began recording tornadoes in 1950. At this longitude in North America, only two tornado events have led to a comparable or greater number of fatalities in the last 125 years: the Worcester, Massachusetts tornado on 9 June 1953 and the Appalachians tornado outbreak on 23 June 1944. The number of violent F4/F5 tornadoes outnumbers all but four tornado events since 1950 (1974 Super Outbreak, 1965 Plam Sunday Tornado Outbreak, 2011 Super Outbreak, 1952 Tornado Outbreak), and its death toll exceeds everything else in the period between the two massive outbreaks of April 1974 and April 2011. It is also one of the only known North American tornado events to produce significant tornadoes in both the U.S. and Canada on the same day (Berrington 2015). The first tornadoes of the day occurred in southeastern Ontario, Canada with 13 tornadoes impacting the region resulting in twelve fatalities with the deadliest being the Barrie, Ontario tornado (eight fatalities). Metrologically, this was a classic tornado setup in a not-so-classic tornado region, physically and socially. In fact, a survey conducted in five contiguous counties in northwestern Pennsylvania found that only 34% of those seriously injured during the 1985 outbreak knew the difference between a tornado warning and a tornado watch [Center for Disease Control (CDC)) 1986]. Perhaps the greatest lesson learned from this outbreak was dispelling the myth shared amongst many Pennsylvanians that tornadoes, or at least destructive ones, cannot occur in the Commonwealth as the hills and mountains provide protection. Quoting the original 1985 survey report produced by the National Oceanic and Atmospheric Administration (NOAA1985):

Table 1. Summary of all tornadoes that impacted Pennsylvania on May 31, 1985. “FATAL” represents the number of fatalities.

No.	Tornado	Start/finish	Fujita	Time (PM, EDT)	Length (miles)	Width (yards)	Area (SQ MI)	Injury	Fatal
334	Albion, PA	OH/PA	4	4:59 − 5:17	14	400	3.18	82	12
336	Linesville, PA	PA	2	5:10 − 5:15	4	200	0.45	0	1
337	Atlantic, PA	OH/PA	4	5:17 − 6:30	56	350	11.14	125	16
338	Saegertown, PA	PA	3	5:23 − 5:55	23	277	3.62	0	0
339	Corry, PA	PA/NY	4	5:25 − 5:55	28	440	7.00	12	0
340	Dorset, OH	OH/PA	2	5:28 − 5:41	10	220	1.25	15	0
341	Centerville, PA	PA	3	6:12 − 6:23	8	427	1.94	10	2
343	Wheatland, PA	OH/PA	5	6:30 − 7:35	47	450	12.02	310	18
344	Tionesta, PA	PA	4	6:30 − 7:10	29	800	13.18	30	7
345	Thompson Run, PA	PA	1	6:30 − 6:37	5	130	0.37	0	0
346	Lamont, PA	PA	2	6:50 − 7:35	19	300	3.24	0	0
349	Tidioute, PA	PA	3	7:30 − 7:55	17	800	7.73	8	0
350	Moshannon State Forest, PA	PA	4	7:35 − 9:00	69	3330	130.55	1	0
353	Big Bend, PA	PA	2	7:54 − 8:03	6	150	0.51	2	0
354	Emlenton, PA	PA	0	7:56 − 8:03	5	300	0.85	0	0
355	Kane, PA	PA	4	8:00 −8:40	29	1000	16.48	40	4
356	Big Beaver, PA	PA	3	8:10 − 9:05	39	243	5.38	120	9
359	Watsontown, PA	PA	3	9:25 − 10:15	19	910	9.82	60	6
360	Penn Run, PA	PA	0	9:53 − 10:02	6	27	0.09	0	0
362	Hollenback Township, PA	PA	1	10:45 − 11:00	11	530	3.31	0	0
363	Unknown, PA	PA	1	12:05A − ?	0	17	0.00	0	0
	TOTAL							815	75

Perhaps the lesson to be learned from the 1985 outbreak is that under the proper atmospheric conditions, major tornadoes can occur irrespective of the location or terrain. In conclusion, given the rarity of the event on a national basis and even more so on a regional basis, it is impressive that the death toll wasn’t worse (p. 9).

What if these atmospheric conditions were to occur again in the same area? What would the impacts be? Tornado historian Thomas Grazulis (1993) calculated the odds at 1 in 75,000 for the outbreak to occur again in the same geographic region. However, the odds may now be greater as several recent studies suggest a change in tornado frequency toward the eastern U.S. Farney and Dixon (2015) noted a modest increase in annual number of tornado days across all the U.S. from 1960 to 2011 with a higher risk area along the east coast from Florida to Pennsylvania. Regionally, Agee et al. (2016) noted a decrease in (E)F1–(E)F5 tornado counts and days from 1954 to 1983 in Texas and Oklahoma and increases in Tennessee and Alabama from 1984 to 2013. The authors concluded the latter demonstrates a new geographical distribution of tornado activity. Gensini and Brooks (2018) found similar results as they noted a robust upward trend of significant tornado parameters (STP) in portions of the southeast, Midwest, and northeast U.S. from 1979 to 2017. Most recently, Ashley, Haberlie, and Gensini (2023) studied the future of U.S. supercells and projected they would become more numerous in eastern regions, while decreasing in frequency in portions of the Great Plains. They further suggest the potential for more significant tornadoes, hail, and extreme rainfall that coupled with vulnerable societies, may produce disastrous consequences. By its standards, Pennsylvania recently experienced several above-average numbers of tornadoes. It ranks twenty-sixth in total number of U.S. tornadoes from 1950 to 2023 (n = 909), averaging 12.6 tornadoes per year (NWS SPC 2023). The last six-years from 2018 to 2023 had 32, 35, 6, 43, 7, and 24, all ranging EF0-EF2 except for one EF3 in 2021. Historically, 2021, 2019, and 2018 were notable as they rank second through fourth in total number of tornadoes, only behind 1998 with the most at 59.

Time will tell if an increase in tornado days shifts east across the U.S. and whether Pennsylvania’s recent tornado activity is indicative of this or even another 1985 tornado outbreak. In the event the latter comes to fruition, this study conducts a spatial analysis of the 1985 tornado outbreak and its impacts on the landscape and population across two time-periods. This is accomplished by analyzing the types and amounts of land cover, total population, and number of buildings within the paths of all 21 tornadoes that impacted Pennsylvania in May 1985. Specifically, a spatiotemporal analysis is conducted for the time-periods representative of year 1985 and present-day to determine changes in these geographies. Ultimately, the question asked by this research is how have changes in land cover, population, and buildings within the 1985 tornado paths influenced Pennsylvania’s tornado impact probability? Like Strader et al. (2017), this study assumes that a greater amount of residential land, people, and buildings within tornado paths translates to a greater tornado disaster probability, and thus an expanding bull’s eye effect (Ashley et al. 2014). This study concludes by conducting correlation and multiple regression analyses between 1985 tornado fatalities and land cover, total population, and building data to estimate the number of present-day fatalities. Except for a mortality study by the CDC (1986) and a tree regrowth analysis (Peterson 2000), there are no studies in the academic literature on the 1985 tornado outbreak. This study will address that shortcoming by providing the first quantification of its impacts, including an analysis on today’s geographic landscape. Overall, this study and its design is motivated in part by previous research (Hall and Ashley 2008; Ashley et al. 2014; Rosencrants and Ashley 2015; Ashley and Strader 2016; Strader et al. 2017; and Strader et al. 2018) that asks similar questions but in completely different study areas.

GIS for spatially analyzing tornado impacts

Several GIS-based studies have analyzed changes in land cover and population to determine regional risk and vulnerability to tornadoes. The primary framework for this particular study is Hall and Ashley (2008). They studied a six-county region surrounding the Chicago, Illinois urban core to determine how urban sprawl impacts vulnerability to tornadoes. By analyzing changes in population (1990 to 2000), land use, housing units, and the value of homes along the suburban fringe of Chicago, they concluded more people and homes are at risk for tornado impacts due to suburban and exurban sprawl. Ashley et at. (2014) expanded on this study to evaluate the concept of the “expanding bull’s-eye effect” in the Chicago region. It’s defined as follows:

This effect argues that targets – that is, humans and their possessions – of geophysical hazards are enlarging as populations grow and spread. Consequently, it is not solely the population magnitude that is important in creating disaster potential; rather, it is how the population, and its affiliated built environment, is distributed across space that determines how the underlying disaster components of risk and vulnerability are realized (Ashley et al. 2014, 186).

They overlaid historical census data as a gridded network from the years 1990, 2000, and 2010 with known and hypothetical tornado paths to produce a set of tornado disaster scenarios. The region’s population grew by 21%, outpaced by a 47.4% increase in housing units. Thus, potential tornado losses would be greater for insured and uninsured housing damages than human fatalities. Ultimately the results confirm that the population and their housing continue to geographically expand outward, increasingly becoming a hazard target known as the expanding bull’s eye effect.

Rosencrants and Ashley (2015) analyzed how sprawling developments increased tornado hazard exposure in the five metropolitan statistical areas (MSAs) of Atlanta, GA; Chicago, IL; Dallas/Fort Worth, TX; Oklahoma City, OK; and St. Louis, MO. Each MSA contains regions characterized as rural, exurban, suburban, and urban developments. The study uses census tract (block) data on population, housing units, and number of households for every decennial census from 1960 to 2010. This data is overlaid with 13 hypothetical tornado paths in each MSA to simulate possible disaster scenarios and determine changes in tornado exposure over the time period. Attributes of the 22 May 2011 Joplin, Missouri EF5 tornado were used to generate various hypothetical paths that were simulated throughout each MSA. Results indicate that each MSA had growth in population and housing units from 1960 to 2010, with the greatest relative increases in Atlanta and Dallas and the smallest in Chicago and St. Louis. Only four of the 80 total counties in the MSAs had a decrease in population (St. Louis and Chicago). Each contained regions classified as urban, thus illustrating the declining population in urban cities and growth outside central cities. Further, all five MSAs experienced increasing total area classified as exurban and suburban, while decreasing in area classified as rural. The results confirm the expanding bull’s eye effect as larger areas of metropolitan regions are becoming exurban and suburban, creating new areas of people and housing that could be affected by all facets of severe storm hazards.

Ashley and Strader (2016) analyzed the changing relationship between housing units and land use in relation to the occurrences of all EF1-EF5 tornadoes over a 60-year period from 1950 to 2010. Specifically, they tallied the number of housing units and land use that intersected each EF1+ tornado to calculate tornado exposure and the probability-risk of being impacted by a tornado based on these past occurrences. The analysis was conducted over 2 national and 4 regional scales. The former are the conterminous United States (US) and the area east of the Continental Divide (CD), which accounts for 93% of all recorded EF1+ tornadoes. The latter are defined as the high plains (HP), central plains (CP), midsouth (MS), and the Midwest (MW). Results indicate that all regions analyzed have experienced greater than 300% growth in probabilistic risk–exposure values since 1950. This is primarily due to the 376.9% increase in housing units since 1950 in the CD region. Most of this growth in housing occurred on exurban and suburban land use, primarily at the expense of rural land, leading to a notable spread in developed land and its people at risk of hazards – the expanding bull’s-eye effect.

Strader et al. (2017) continued research on the HP, CP, MS, and MW regions with an eye on the future. They researched tornado exposure by evaluating how human development has amplified disaster potential, specifically, how changes in housing units and developed land may influence disaster frequency and consequences in the future. Results illustrate that although the MW region contains the greatest built-environment exposure and the CP experiences the most significant tornadoes, the MS contains the greatest tornado disaster potential. This is due to its elevated tornado risk and accelerated growth in developed land area that characterizes the MS region. Annual tornado impacts for all regions are projected to nearly double during the twenty-first century, signifying the potential for greater tornado disaster potential in the future. Last, Strader et al. (2018) analyzed how different types and the spatial characteristics of land use influence tornado impact magnitude and probability in the CP region. It was chosen because of its large proportion of rural land surrounding densely populated metropolitan areas. They concluded that tornado impact magnitude and probability are strongly influenced by the spatial character of the residential built environment, particularly communities with greater a development of sprawl.

Data and methods

Forty-three tornadoes occurred during the 31 May 1985 tornado outbreak. The study area for this research are the 21 tornadoes that impacted Pennsylvania; four began in Ohio and ended in Pennsylvania, one began in Pennsylvania and ended in New York, with the remaining 16 beginning and ending in Pennsylvania. This study seeks to quantify their impacts across two time-periods on land cover, population, and building structures to determine if they exhibit characteristics of the expanding bull’s eye effect. Careful consideration was given in selecting geospatial datasets to represent these variables to allow for uniform comparison between the time-periods of analysis. All spatial analyses were conducted in ArcGIS Pro.

Tornado paths

Data representing tornado paths were obtained from NOAA’s NWS SPC Severe Weather GIS (SVRGIS 2023) download page (http://www.spc.noaa.gov/gis/svrgis/). The file format is shapefile (.shp) with the dataset representing each tornadoes straight-line path from start to finish. This study area’s 21 tornadoes were plotted in ArcGIS Pro and their width mapped using the buffer tool. This was done to ultimately determine the amount and type of land cover, population characteristics, and number of buildings within each tornado path.

Land cover data

Land cover data representative of the 1985 and present-day time-periods were obtained from the Multi-Resolution Land Characteristics (MRLC) Consortium. They are the only organization that provides comparable land cover data for the two time periods. Specifically, 1992 and 2019 National Land Cover Dataset (NLCD) were utilized in this research. The 1992 NLCD [United States Geological Survey (USGS)) 2000] is the earliest data-set available from the MRLC and thus most closely represents the 1985 time-period (https://www.sciencebase.gov/catalog/item/5dfbcd38e4b0ff479b8c459f). The 2019 NLCD (Dewitz and USGS 2021) is the most recent available (https://www.mrlc.gov/data/nlcd-2019-land-cover-conus). NLCD 1992 is a 21-class legend primarily based on the unsupervised classification of Landsat Thematic Mapper (TM) circa 1990 satellite data and has a spatial resolution of 30 meters (Vogelmann et al. 2001). Other ancillary data sources were used to create the data: topography, census, agricultural statistics, soil characteristics, and other types of land cover and wetland maps. NLCD 2019 is a 16-class legend based on the analysis of Landsat data at a resolution of 30 meters (Yang et al. 2018). Both legends represent land cover according to a modified Anderson Level II (descriptive) land-use and land-cover (LULC) classification system (Anderson et al. 1976). The difference in the number of legend classes between the two exists because post-1992 mapping methods consolidated some Anderson Level II classes while dividing others into higher resolution classes to meet the thematic needs of NLCD (Wickham et al. 2004). Particularly noteworthy for this study is the number of urban categories represented by each. NLCD 1992 legend represents urban or built-up land with three classes: Low Intensity Residential, High Intensity Residential, and Commercial/Industrial/Transportation. NLCD 2019 uses four classes: Developed Open Space, Developed Low Intensity, Developed Medium Intensity, and Developed High Intensity. There are other differences where certain classes are represented on NLCD 1992, like Transitional Barren, but not on NLCD 2019 and vice versa. These differences prevent direct comparison of land cover changes at their Anderson II Levels. Thus, the legend classes for each were changed from Anderson Level II (descriptive) to Anderson Level I (general) by combining them into 8 broader classes: Open Water, Urban, Barren Land, Forest, Grass/Shrub, Agriculture, and Wetlands. The tornado paths were plotted overtop each land cover dataset. An ArcGIS Pro clip created a single layer that represents land cover solely within the path of each tornado. The type and acreage of land cover within the clipped tornado paths was calculated for each time period with urban being the focus of this study as it represents low, medium, and high intensity residential. This was done to determine the change in urban land cover within the tornado paths and how it relates to the expanding bull’s eye effect.

Census data

Census data was obtained from both the National Historical Geographic Information System (NHGIS) (https://www.nhgis.org/) and the U.S. Census Bureau (https://data.census.gov/table). NHGIS provides summary demographic data and GIS boundary files for years 1790 to present-day across census geographic levels of states, counties, tracts, block groups, and blocks Complete demographic datasets at the tract, block group, and block level representing year 1980 are not available across Pennsylvania, Ohio, and New York. Further, complete demographic datasets are also not available at the census block level for 1990 but are available at the census tract and block group levels. Thus, this research used 1990 (Manson et al. 2022) and 2020 (U.S. Census Bureau 2023) census block group data to represent the characteristics of the population. The following categories were analyzed: total population, number of households, age under 5, age 5–9, age 10–19, age 20–29, age 30–39, age 40–49, age 50–59, age 60–69, age 70–79, age 80 and over. This data was used to determine the demographic characteristics of the population within the 21 tornado paths and how changes up to present-day, coupled with changes in land cover, relate to the expanding bull’s eye effect.

Overlaying the tornado tracks with the census block groups can cause entire block groups to be selected even though only a small portion is within the tornado path. Demographic characteristics would therefore be overestimated (Schlossberg 2003; Hall and Ashley 2008) through the modifiable areal unit problem (MAUP) (Gehlke and Biehl 1934; Openshaw and Taylor 1979). This study chose to compensate for the overestimation by adjusting population data according to both the proportion of the block group within tornado’s path and its land cover characteristics. Specifically, for portions of block groups within a tornado’s path, the percentage of urban land cover it contains of the whole block group was calculated. This percentage was then multiplied against the block groups total population. For example, a census block group contains 50 total acres of urban land cover and has a total population of 500. A portion of it within a tornado’s path contains 10 urban acres, 20% of that block group’s total urban land cover. This percentage was multiplied against the block group’s total population (0.2 × 500) to get an estimate of 100 people in the tornado’s path. This calculation was performed to estimate all demographic categories. If a tornado path intersected a portion of a block group with no urban land cover, it was assumed no one was affected and no demographics were calculated. This method is similar to areal weighting (AW) utilized by Ashley et al. (2014) in quantifying demographics characteristics within tornado paths of the Chicago metropolitan region. It assumes an equal distribution of the population across urban land cover. As previously stated, the alternative is overestimating the population like Hall and Ashley (2008). Their method calculates the demographic characteristics of those affected by summing the attribute values for all block groups within or intersected by each tornado path, even if the path merely clipped a small proportion of the block enumeration. Thus, this study on the May 1985 tornadoes favors AW in calculating demographics characteristics of the population.

Buildings data

This study utilized Digital Raster Graphics (DRGs) – scanned 1:24,000 United States Geological Survey (USGS) topographic maps – to create a buildings geospatial dataset representing the 1985 time period. DRGs depict the locations of many geographic features including elevation, roads, railroads, rivers, streams, lakes, boundaries, and of significance to this study, the locations of buildings. A total of 102 DRGs were obtained from the following GIS digital data clearinghouses: The Pennsylvania Spatial Data Access Center (http://www.pasda.psu.edu/) (PASDA 2022a), the Ohio Geographically Referenced Information Program (https://gis1.oit.ohio.gov/geodatadownload/) (OGRIP 2022a), and the New York State GIS Clearinghouse (https://data.gis.ny.gov/) [New York State (NYS) GIS 2022a]. Buildings located within each tornado path were “heads-up” digitized from DRGs to create the 1985 buildings dataset. The main limitation of DRGs is they do not depict the locations of buildings for only year 1985. They were created between 1954 and 1971 with each undergoing a photo revision to improve their accuracy sometime between 1970 and 1990.

Data representing present-day building locations were obtained from the same three GIS digital data clearinghouses: PASDA (2022b), OGRIP (2022b), and NYS GIS (2022b). Each was created by geocoding known buildings or structures according to their physical address. The PASDA data represents buildings from 2007, OGRIP from 2020, and NYS GIS from 2022. They were standardized to a single-time period by overlaying each with 2023 World Imagery basemap (https://www.arcgis.com/home/item.html?id=10df2279f9684e4a9f6a7f08febac2a9) in ArcGIS Pro and performing heads-up digitizing to add or remove buildings. A clip and spatial join calculated the total number of 1985 and 2023 buildings within the tornado paths. Figure 2 illustrates a portion of the Watsontown tornado overlaid all geospatial datasets. To summarize, this study quantifies the effects of 21 tornadoes across time periods representing 1985 and present-day using GIS methods applied to several geospatial datasets as outlined in Figure 3. These results are then applied to correlation and regression analyses to estimate the number of fatalities for present-day.

Figure 2. Portion of 359-Watsontown tornado overlaid and intersecting: A. land cover data B. buildings data C. census blocks groups and D. urban land cover (grey) within each census block group..

Figure 3. Synopsis of Data and Methods.

Results

1985 impacts

Ninety-six percent of land cover within all 21 tornado paths is Forest and Agriculture as the former represents the most at 86.3% (Table 2). Urban represents 1.3% (1,949.1 acres) and contains an estimated population of 12,778 (Table 3), most being young to middle-aged (age 10 to 49). A total of 4,292 buildings were impacted (Table 4). Nine tornadoes caused all the fatalities with eight having a Fujita rating of three or higher. Three tornadoes – Wheatland-F5, Atlantic-F4, and Albion-F4 – (hereafter referred to as the “Big-Three”) were responsible for over-half (n = 46, 61%) of the fatalities. Wheatland, the deadliest, at 450 yards wide was on the ground for 47 miles. Of all tornadoes, it overwhelmingly contains the greatest amount of urban land with 1,070.5 acres. Given this amount, it’s fortunate more fatalities did not occur as the estimated total population within its path is 5,612; the highest population of all tornadoes, the majority being middle-aged. Further, this tornado also impacted the greatest number of buildings at 1,188. Atlantic, the second deadliest tornado, contains the second highest population (2,009) but ranks fourth in urban land cover (115.1 acres) and buildings (357). The third and fourth deadliest tornadoes, Albion (12) and Big Beaver (9), also have the third and fourth largest populations but the former has fewer people (1,128) than the latter (1,755). The Albion tornado was only on the ground for 14 miles (tenth longest length) but still affected 354 buildings across 114.8 acres of urban land cover, the fifth highest amounts for both categories. The path of the weaker Big Beaver (F3) tornado was 25 miles longer impacting a total of 254.7 acres of urban land cover (second highest) and 483 buildings (third highest). It has more fatalities than four other F4 tornadoes – Corry, Tionesta, Moshannon State Forest, and Kane – presumably due to the 1,755 people its path, the third highest amount. The Moshannon State Forest tornado had the longest path of 69 miles but caused no fatalities and only one injury. Appropriately named, 96% of its land cover (83,576.9 acres) is forest with less than 1% being urban. However, it impacted an estimated 400 people (seventh highest) and 704 buildings (second highest).

Table 2. Total amount (acres) and type of 1992 Anderson Level I land cover within each tornado path.

No.	Tornado	Fujita	Fatal	Injury	Length	Width	Open water	Urban	Barren land	Forest	Grass/ Shrub	Agriculture	Wetlands
334	Albion	4	12	82	14	400	2.3	114.8	0.0	1,264.9	0.0	657.9	199.1
336	Linesville	2	1	0	4	200	0.8	5.9	0.0	157.8	0.0	137.5	7.7
337	Atlantic	4	16	125	56	350	16.8	115.1	28.2	4,353.6	0.0	2,144.6	114.0
338	Saegertown	3	0	0	23	277	4.7	5.3	0.5	1,198.1	0.0	980.2	24.9
339	Corry	4	0	0	28	440	0.1	3.6	18.8	2,750.9	0.0	1,564.2	109.9
340	Dorset	2	0	15	10	220	3.3	5.5	0.0	332.1	0.0	476.3	139.7
341	Centerville	3	2	0	8	427	1.5	0.7	1.6	265.7	0.0	97.1	0.0
343	Wheatland	5	18	310	47	450	66.9	1,070.5	11.6	3,873.2	0.0	1,930.2	469.7
344	Tionesta	4	7	30	29	800	27.6	59.7	54.2	10,231.5	0.0	265.8	9.7
345	Thompson Run	1	0	0	5	130	0.0	0.0	0.2	254.1	0.0	0.9	0.0
346	Lamont	2	0	0	19	300	0.0	6.2	10.2	1,761.4	0.0	84.8	0.0
349	Tidioute	3	0	8	17	800	17.5	5.1	125.1	4,711.7	0.0	154.2	87.9
350	Moshannon State Forest	4	0	0	69	3330	406.7	54.4	703.5	83,576.9	0.0	1,723.2	165.9
353	Big Bend	2	0	1	6	150	25.1	0.2	0.3	263.0	0.0	64.0	0.0
354	Emlenton	0	0	0	5	300	23.6	29.9	3.7	521.2	0.0	122.5	0.0
355	Kane	4	4	40	29	1000	127.8	62.8	115.0	9,822.1	0.0	179.1	12.4
356	Big Beaver	3	9	120	39	243	7.3	254.7	34.7	1,850.1	0.0	1,271.3	0.0
359	Watsontown	4	6	60	19	910	75.4	33.2	0.0	2,292.6	0.0	3,411.4	5.4
360	Penn Run	0	0	0	6	27	0.0	0.1	0.2	15.8	0.0	6.8	0.0
362	Hollenback Township	1	0	0	11	530	17.9	121.4	19.3	1,755.3	0.0	315.6	30.5
363	Unknown	1	0	0	0	17	0.0	0.0	0.0	0.1	0.0	0.0	0.0
	TOTAL	—	75	791	Miles	Yards	825.3	1,949.1	1,127.1	131,252.1	0.0	15,587.6	1,376.8
							0.5%	1.3%	0.7%	86.3%	0.0%	10.2%	0.9%

Table 3. Demographic characteristics of the 1985 population within each tornado path.

No.	Tornado	Pop	HHs	Under 5	5–9	10–19	20–29	30–39	40–49	50–59	60–69	70–79	80 over
334	Albion	1,128	430	79	90	177	155	151	138	108	115	78	37
336	Linesville	50	20	2	3	6	5	7	7	6	6	6	2
337	Atlantic	2,008	773	132	133	275	255	291	262	214	219	150	77
338	Saegertown	156	56	11	12	25	19	25	22	17	15	8	2
339	Corry	91	29	7	8	16	12	15	13	9	6	4	1
340	Dorset	22	7	2	2	4	3	4	2	2	2	1	0
341	Centerville	22	8	2	2	4	3	3	3	2	2	1	0
343	Wheatland	5,612	2,175	376	372	764	707	841	676	554	651	475	196
344	Tionesta	200	76	13	14	27	22	31	29	23	24	13	4
345	Thompson Run	0	0	0	0	0	0	0	0	0	0	0	0
346	Lamont	11	4	1	1	2	1	2	1	1	1	1	0
349	Tidioute	48	19	2	4	6	5	7	7	5	7	3	2
350	Moshannon State Forest	402	151	26	26	59	59	65	50	42	40	25	10
353	Big Bend	1	1	0	0	0	0	0	1	0	0	0	0
354	Emlenton	73	29	4	4	10	9	10	10	8	9	6	3
355	Kane	479	179	32	35	74	63	66	64	48	48	37	12
356	Big Beaver	1,755	636	117	128	254	243	301	212	185	177	102	36
359	Watsontown	271	88	21	16	47	38	44	36	25	23	13	8
360	Penn Run	2	1	0	0	0	1	1	0	0	0	0	0
362	Hollenback Township	447	156	39	38	63	48	89	61	36	37	27	9
363	Unknown	0	0	0	0	0	0	0	0	0	0	0	0
	TOTAL	12,778	4,838	866	888	1,813	1,648	1,953	1,594	1,285	1,382	950	399
				6.8%	6.9%	14.2%	12.9%	15.3%	12.5%	10.1%	10.8%	7.4%	3.1%

Table 4. Number of buildings within each tornado path for the 1985 and 2023 (present-day) time periods.

No.	Tornado	Start/ finish	Fujita	Fatal	Injury	Length	Width	1985 buildings	2023 buildings	1985 total	2023 total	Absolute change	Percent change
334	Albion	OH/PA	4	12	82	14	400	2/352	4/413	354	417	63	17.8%
336	Linesville	PA	2	1	0	4	200	23	28	23	28	5	21.7%
337	Atlantic	OH/PA	4	16	125	56	350	0/357	0/798	357	798	441	123.5%
338	Saegertown	PA	3	2	0	23	277	64	169	64	176	112	175.0%
339	Corry	PA/NY	4	0	0	28	440	27/20	41/58	47	99	52	110.6%
340	Dorset	OH/PA	2	0	15	10	220	45/3	44/3	48	47	−1	−2.1%
341	Centerville	PA	3	0	0	8	427	4	24	4	24	20	500.0%
343	Wheatland	OH/PA	5	18	310	47	450	899/289	1,329/1,022	1,188	2,351	1,163	97.9%
344	Tionesta	PA	4	7	30	29	800	150	332	150	332	182	121.3%
345	Thompson Run	PA	1	0	0	5	130	2	2	2	2	0	0.0%
346	Lamont	PA	2	0	0	19	300	23	41	23	41	18	78.3%
349	Tidioute	PA	3	0	8	17	800	114	298	114	298	184	161.4%
350	Moshannon State Forest	PA	4	0	0	69	3330	704	1,316	704	1,316	612	86.9%
353	Big Bend	PA	2	0	1	6	150	6	32	6	32	26	433.3%
354	Emlenton	PA	0	0	0	5	300	34	75	34	75	41	120.6%
355	Kane	PA	4	4	40	29	1000	332	631	332	631	299	90.1%
356	Big Beaver	PA	3	9	120	39	243	483	801	483	801	318	65.8%
359	Watsontown	PA	4	6	60	19	910	204	669	204	669	465	227.9%
360	Penn Run	PA	0	0	0	6	27	0	2	0	2	2	N/A
362	Hollenback Township	PA	1	0	0	11	530	155	304	155	304	149	96.1%
363	Unknown	PA	1	0	0	0	17	0	0	0	0	0	N/A
	TOTAL	—	—	75	791	Miles	Yards	—	—	4,292	8,443	4,151	96.7%

The 29-mile-long Tionesta tornado was the fifth deadliest (7) despite only affecting 59.7 acres of urban land cover (seventh highest amount) that contained 200 people and 150 buildings, both the ninth highest amounts of all tornadoes. The Watsontown tornado is the sixth deadliest tornado, affecting 33.2 acres of urban land cover, 271 people, and 204 buildings across 19 miles (ninth longest path). Despite being rated F4, the Kane tornado is only the seventh deadliest tornado with 4 fatalities. It traveled 29 miles (tied fifth longest path) and impacted 480 people (fifth highest) and 332 buildings (sixth highest) across 62.8 acres (sixth highest). The two remaining tornadoes that caused fatalities, Centerville (2) and Linesville (1), each had paths less than 10 miles and impacted less than 50 people. However, it is worth noting that the Centerville tornado impacted less than 1 acre of urban land cover with an estimated population of 24 people and four buildings. The two fatalities and 10 injuries from this tornado indicate that half the population in its path were injured or fatally injured. Of the 12 tornadoes that did not result in any fatalities, the aforementioned Moshannon State Forest tornado is the most interesting for reasons already stated. Two other tornadoes are of interest. The Corry tornado was rated F4 across a 28-mile path. Fortunately, it only impacted 3.6 acres of urban land cover (sixteenth highest) with a population of 92 (eleventh highest) and 47 buildings (thirteenth highest). The 11-mile, F1 Hollenback Township tornado impacted 121.4 acres of urban land cover – the third highest amount. As a result, an estimated 449 people (sixth highest) and 155 buildings (eighth highest) were impacted but resulted in zero fatalities or injuries.

Changes in land cover, population, and buildings for present-day

Forest and agriculture remain the highest proportion of land cover in the tornado paths (Table 5) but at a lower overall percentage (89.9%) as urban represents a greater amount (5.9%) compared to 1992. Four of the seven land cover categories increased acreage since 1992. The largest absolute and percentage increases were urban at 6,998 additional acres, a 359% increase (Table 6). This was at the expense of forest, agriculture, and barren land. Forest and Agriculture had the largest absolute decreases (-5,835.3, −3,702.90 acres) with the latter representing the largest percentage decrease (-82.5%). A 96.7% increase in the number of buildings nearly doubled the 1985 amount, bringing the total to 8,443 (Table 4) with every tornado exhibiting an increase. Unlike the increases in urban and buildings, the total population decreased from 12,778 in 1990 to 11,791 in 2020 (Table 7), a loss of 990 people (-7.7%) (Table 8). However, this decrease was not spatially uniform across all tornadoes. Nine tornadoes lost population, eleven gained, and one had no change. Losses occurred in all age cohorts under 50 years with increases in all others above 50. As a result, those aged 50 to 59 and 60 to 69 now represent the highest percentage, signifying a shift to a more elderly population in tornado paths.

Table 5. Total amount (acres) and type of 2019 Anderson Level I land cover within each tornado path.

No.	Tornado	Fujita	Fatal	Injury	Length	Width	Open water	Urban	Barren land	Forest	Grass/ Shrub	Agriculture	Wetlands
334	Albion	4	12	82	14	400	0.0	274.3	0.0	904.5	2.3	477.0	586.0
336	Linesville	2	1	0	4	200	0	27.3	0.0	161.4	11.7	93.4	16.0
337	Atlantic	4	16	125	56	350	50.2	516.5	0.9	4,440.9	53.1	1,661.6	248.6
338	Saegertown	3	0	0	23	277	11.7	179.0	0.7	1,122.3	17.3	749.4	146.8
339	Corry	4	0	0	28	440	13.7	196.7	10.2	2,645.6	21.8	1,237.5	390.0
340	Dorset	2	0	15	10	220	0.6	54.1	0.4	360.7	7.7	428.0	107.4
341	Centerville	3	2	0	8	427	4.6	20.8	0.0	210.6	1.0	86.2	44.1
343	Wheatland	5	18	310	47	450	61.4	1,866.1	49.4	3,739.7	194.7	1,195.9	432.9
344	Tionesta	4	7	30	29	800	87.4	453.6	6.7	9,620.0	93.0	199.6	249.0
345	Thompson Run	1	0	0	5	130	0.0	3.0	0.0	249.6	0.4	2.3	0.0
346	Lamont	2	0	0	19	300	0.0	77.6	0.0	1,339.4	19.5	44.9	57.6
349	Tidioute	3	0	8	17	800	31.3	265.2	8.3	4,396.0	75.4	224.7	107.2
350	Moshannon State Forest	4	0	0	69	3330	470.4	2,883.4	71.5	80,532.7	809.8	1,476.9	846.6
353	Big Bend	2	0	1	6	150	39.7	14.8	0.0	243.0	6.2	49.1	0.1
354	Emlenton	0	0	0	5	300	58.6	114.0	0.0	401.0	17.6	108.8	1.3
355	Kane	4	4	40	29	1000	127.3	518.8	33.9	8,911.5	212.5	274.5	268.3
356	Big Beaver	3	9	120	39	243	20.2	772.8	3.6	1,713.1	58.3	918.2	7.8
359	Watsontown	4	6	60	19	910	90.7	422.4	9.7	2,858.9	52.3	2,311.5	79.3
360	Penn Run	0	0	0	6	27	0.3	0.9	0.0	13.7	0.0	8.0	0.0
362	Hollenback Township	1	0	0	11	530	6.9	285.8	2.0	1,552.2	54.0	337.2	23.9
363	Unknown	1	0	0	0	17	0.0	0.0	0.0	0.0	0.0	0.0	0.1
	TOTAL	—	75	791	Miles	Yards	1,075.0	8,947.1	197.3	125,416.8	1708.6	11,884.7	3,613.0
							0.7%	5.9%	0.1%	82.1%	1.1%	7.8%	2.4%

Table 6. Table 6. Absolute (acres) and percentage change in land cover within each tornado path from 1992 to 2019.

No.	Tornado	Open water		Urban		Barren land		Forest		Grass/Shrub		Agriculture		Wetlands
334	Albion	−2.3	−100.0%	159.5	139%	0.0	- - - - -	−360.4	−28.5%	2.3	- - - - -	−180.9	−27.5%	386.9	194.3%
336	Linesville	−0.8	−100.0%	21.4	363%	0.0	- - - - -	3.6	2.3%	11.7	- - - - -	−44.1	−32.1%	8.3	107.8%
337	Atlantic	33.4	198.8%	401.4	349%	−27.3	−96.8%	87.3	2.0%	53.1	- - - - -	−483.0	−22.5%	134.6	118.1%
338	Saegertown	7.0	148.9%	173.7	3277%	0.2	40.0%	−75.8	−6.3%	17.3	- - - - -	−230.8	−23.5%	121.9	489.6%
339	Corry	13.6	13600.0%	193.1	5364%	−8.6	−45.7%	−105.3	−3.8%	21.8	- - - - -	−326.7	−20.9%	280.1	254.9%
340	Dorset	−2.7	−81.8%	48.6	884%	0.4	- - - - -	28.6	8.6%	7.7	- - - - -	−48.3	−10.1%	−32.3	−23.1%
341	Centerville	3.1	206.7%	20.1	2871%	−1.6	−100.0%	−55.1	−20.7%	1.0	- - - - -	−10.9	−11.2%	44.1	- - - - -
343	Wheatland	−5.5	−8.2%	795.6	74%	37.8	325.9%	−133.5	−3.4%	194.7	- - - - -	−734.3	−38.0%	−36.8	−7.8%
344	Tionesta	59.8	216.7%	393.9	660%	−47.5	−87.6%	−611.5	−6.0%	93.0	- - - - -	−66.2	−24.9%	239.3	2467.0%
345	Thompson Run	0.0	- - - - -	3.0	- - - - -	−0.2	−100.0%	−4.5	−1.8%	0.4	- - - - -	1.4	155.6%	0.0	- - - - -
346	Lamont	0.0	- - - - -	71.4	1152%	−10.2	−100.0%	−422.0	−24.0%	19.5	- - - - -	−39.9	−47.1%	57.6	- - - - -
349	Tidioute	13.8	78.9%	260.1	5100%	−116.8	−93.4%	−315.7	−6.7%	75.4	- - - - -	70.5	45.7%	19.3	22.0%
350	Moshannon State Forest	63.7	15.7%	2,829.0	5200%	−632.0	−89.8%	−3,044.2	−3.6%	809.8	- - - - -	−246.3	−14.3%	680.7	410.3%
353	Big Bend	14.6	58.2%	14.6	7300%	−0.3	−100.0%	−20.0	−7.6%	6.2	- - - - -	−14.9	−23.3%	0.1	- - - - -
354	Emlenton	35.0	148.3%	84.1	281%	−3.7	−100.0%	−120.2	−23.1%	17.6	- - - - -	−13.7	−11.2%	1.3	- - - - -
355	Kane	−0.5	−0.4%	456.0	726%	−81.1	−70.5%	−910.6	−9.3%	212.5	- - - - -	95.4	53.3%	255.9	2063.7%
356	Big Beaver	12.9	176.7%	518.1	203%	−31.1	−89.6%	−137.0	−7.4%	58.3	- - - - -	−353.1	−27.8%	7.8	- - - - -
359	Watsontown	15.3	20.3%	389.2	1172%	9.7	- - - - -	566.3	24.7%	52.3	- - - - -	−1,099.9	−32.2%	73.9	1368.5%
360	Penn Run	0.3	- - - - -	0.8	800%	−0.2	−100.0%	−2.1	−13.3%	0.0	- - - - -	1.2	17.6%	0.0	- - - - -
362	Hollenback Township	−11.0	−61.5%	164.4	135%	−17.3	−89.6%	−203.1	−11.6%	54.0	- - - - -	21.6	6.8%	−6.6	−21.6%
363	Unknown	0.0	- - - - -	0.0	- - - - -	0.0	- - - - -	−0.1	−100.0%	0.0	- - - - -	0.0	- - - - -	0.1	- - - - -
	TOTAL	249.7	30.3%	6,998.0	359%	−929.8	−82.5%	−5,835.3	−4.4%	1,708.6	- - - - -	−3,702.9	−23.8%	2,236.2	162.4%

Table 7. Demographic characteristics of the 2020 population within each tornado path.

No.	Tornado	Pop	HHs	Under 5	5–9	10–19	20–29	30–39	40–49	50–59	60–69	70–79	80 over
334	Albion	705	300	34	46	92	63	66	96	116	108	57	26
336	Linesville	33	14	2	2	4	3	4	3	5	6	3	2
337	Atlantic	808	358	37	33	91	90	59	111	121	112	99	55
338	Saegertown	226	88	14	11	30	35	18	27	33	30	17	11
339	Corry	246	91	12	19	36	23	25	26	41	40	18	6
340	Dorset	65	22	5	7	10	5	11	5	8	6	5	2
341	Centerville	26	8	3	3	4	3	3	2	3	3	2	1
343	Wheatland	4,284	1,958	232	246	549	474	432	531	609	530	412	267
344	Tionesta	173	82	3	2	14	14	13	16	30	36	32	13
345	Thompson Run	4	2	0	0	0	0	0	1	1	0	0	1
346	Lamont	29	13	1	2	2	2	3	3	4	5	3	3
349	Tidioute	84	41	3	3	6	6	10	10	13	20	10	3
350	Moshannon State Forest	1,977	802	112	99	202	216	165	241	393	285	167	97
353	Big Bend	9	4	0	0	1	1	1	1	2	2	1	0
354	Emlenton	71	32	2	3	8	7	6	8	14	12	8	4
355	Kane	171	78	8	10	16	12	21	17	25	32	22	9
356	Big Beaver	1,428	589	71	89	156	127	175	184	200	216	137	73
359	Watsontown	916	152	33	28	54	120	208	189	143	77	47	15
360	Penn Run	1	0	0	0	0	0	0	0	1	0	0	0
362	Hollenback Township	536	221	24	39	57	59	61	58	90	91	32	25
363	Unknown	0	0	0	0	0	0	0	0	0	0	0	0
	TOTAL	11,791	4,856	597	645	1,332	1,262	1,280	1,530	1,852	1,611	1,072	612
				5.1%	5.5%	11.3%	10.7%	10.9%	13.0%	15.7%	13.7%	9.1%	5.2%

Table 8. Absolute and percent change in the population from 1990 to 2020 within each tornado path.

No.	Tornado	Pop	HHs	Under 5	5–9	10–19	20–29	30–39	40–49	50–59	60–69	70–79	80 over
334	Albion (Absolute Change)	−423	−130	−45	−44	−85	−92	−85	−42	8	−7	−21	−11
	(Percent Change)	−37%	−30%	−57%	−49%	−48%	−59%	−56%	−30%	8%	−6%	−27%	−31%
336	Linesville	−17	−6	0	−1	−2	−2	−3	−4	−1	0	−3	0
		−34%	−28%	−11%	−24%	−32%	−38%	−46%	−61%	−21%	1%	−52%	−16%
337	Atlantic	−1,200	−415	−95	−100	−184	−165	−232	−151	−93	−107	−51	−22
		−60%	−54%	−72%	−75%	−67%	−65%	−80%	−58%	−44%	−49%	−34%	−29%
338	Saegertown	70	32	3	−1	5	16	−7	5	16	15	9	9
		45%	57%	23%	−4%	21%	84%	−30%	22%	94%	100%	107%	462%
339	Corry	155	62	5	11	20	11	10	13	32	34	14	5
		170%	215%	75%	139%	123%	95%	64%	103%	354%	560%	351%	525%
340	Dorset	43	15	3	5	6	2	7	3	6	4	4	2
		195%	211%	173%	257%	154%	63%	168%	175%	303%	217%	408%	- - - -
341	Centerville	4	0	1	1	0	0	0	−1	1	1	1	1
		18%	6%	37%	53%	−3%	11%	−5%	−35%	48%	40%	76%	- - - -
343	Wheatland	−1,328	−217	−144	−126	−215	−233	−409	−145	55	−121	−63	71
		−24%	−10%	−38%	−34%	−28%	−33%	−49%	−22%	10%	−19%	−13%	36%
344	Tionesta	−27	6	−10	−12	−13	−8	−18	−13	7	12	19	9
		−14%	8%	−75%	−84%	−48%	−38%	−59%	−45%	32%	49%	149%	215%
345	Thompson Run	4	2	0	0	0	0	0	1	1	0	0	1
		- - - -	- - - -	- - - - -	- - - -	- - - -	- - - -	- - - -	- - - -	- - - -	- - - -	- - - -	- - - -
346	Lamont	18	9	0	1	0	1	1	2	3	4	2	3
		160%	221%	40%	89%	4%	102%	54%	221%	293%	409%	235%	- - - -
349	Tidioute	36	22	1	−1	0	1	3	3	8	13	7	1
		75%	115%	45%	−36%	−5%	20%	47%	47%	166%	190%	230%	35%
350	Moshannon State Forest	1,575	651	86	73	143	157	100	191	351	245	142	87
		392%	431%	329%	282%	242%	266%	155%	382%	835%	612%	568%	873%
353	Big Bend	8	3	0	0	1	1	1	0	2	2	1	0
		792%	299%	- - - - -	- - - - -	- - - - -	- - - - -	- - - - -	3%	- - - -	- - - -	- - - -	- - - -
354	Emlenton	−2	3	−2	−1	−2	−2	−4	−2	6	3	2	1
		−3%	11%	−53%	−20%	−23%	−25%	−43%	−16%	71%	29%	29%	39%
355	Kane	−308	−101	−24	−25	−58	−51	−45	−47	−23	−16	−15	−3
		−64%	−56%	−76%	−71%	−78%	−81%	−68%	−73%	−49%	−34%	−42%	−25%
356	Big Beaver	−327	−47	−46	−39	−98	−116	−126	−28	15	39	35	37
		−19%	−7%	−40%	−30%	−38%	−48%	−42%	−13%	8%	22%	35%	103%
359	Watsontown	645	64	12	12	7	82	164	153	118	54	34	7
		238%	73%	59%	73%	16%	216%	373%	426%	473%	236%	261%	90%
360	Penn Run	−1	−1	0	0	0	−1	−1	0	1	0	0	0
		−42%	−56%	- - - - -	- - - -	- - - -	−88%	−87%	- - - -	- - - -	- - - -	- - - -	- - - -
362	Hollenback Township	89	65	−15	1	−6	11	−28	−3	54	54	5	16
		20%	41%	−37%	2%	−10%	24%	−32%	−5%	150%	146%	19%	175%
363	Unknown	0	0	0	0	0	0	0	0	0	0	0	0
		- - - -	- - - -	- - - - -	- - - -	- - - -	- - - -	- - - -	- - - -	- - - -	- - - -	- - - -	- - - -
Absolute Change in Population		−987	18	−269	−243	−481	−386	−673	−64	567	229	122	213
Percent Change in Population		−7.7%	0.4%	−31.1%	−27.4%	−26.5%	−23.4%	−34.5%	−4.0%	44.1%	16.6%	12.8%	53.4%

Bold represents increases, italics represents decreases. 

Every tornado path had an increase in urban land cover and number of buildings. The lone exception is 363-Unknown-F1 which is excluded from this discussion because its path length is zero miles and contains no urban. One positive observation is that the Big-Three are among those with the smallest percentage increases in urban land: Wheatland (70%), Atlantic (273%), and Albion (136%). They represent the nineteenth, fourteenth, and seventeenth highest percentage increases, but these tornadoes still rank among the highest in total urban land cover: second, fifth, and ninth highest. Wheatland continues to have the most buildings in its path with a 97.9% increase and the largest absolute change of 1,163. The Atlantic tornado more than doubled its total with a 123.5% increase but is still ranked fourth highest in total number of buildings. It’s noteworthy that Atlantic is one of only four tornadoes to experience an increase in forest land cover. Albion has one of the smallest percentage increases (17.8%) with only 63 additional buildings since 1985. Most concerning though is that Noah’s Ark Daycare and Hasson Heights Elementary School are in the paths of the Wheatland and Atlantic tornadoes. Despite their growth in urban and buildings, the Big-Three all experienced population loss. Each experienced the largest absolute change: Wheatland (-1,328), Atlantic (-1,201), and Albion (-423) with the greatest loss in those younger than 40 years old.

The fourth deadliest tornado, Big Beaver, is similar to the Big-Three with a lower percentage increase in urban land cover (157%) but among the largest present-day totals (772.8 acres) – third largest. It also had a 19% decrease in total population but remains the second most populated (1,428) tornado path with the third most buildings (801), one being Buffalo Elementary School. Tionesta, the fifth deadliest tornado, ranks sixth in urban (453.6 acres) but it too lost population but increased buildings. It ranks tenth in population (173) and eighth in buildings (332). Of the remaining tornadoes that caused fatalities, a few changes are noteworthy. Watsontown had a 1,172% increase in urban land, a 238% increase in population, and a 227.9% increase in buildings, which includes the Immaculate Conception School. This is the seventh (422.4 acres), fourth (916), and third highest (669) amounts for present-day. The only other deadly tornado to have an increase in population is Saegertown with a 44% increase bringing the total to 226, supported by an additional 173.7 urban acres and 112 buildings. Arguably, the Moshannon State Forest tornado exhibits the most dramatic changes. It has the largest absolute increase in both urban land cover and total population. It expanded from 54.4 to 2,883.4 acres of urban land cover and from 400 to 1,977 people. The is the largest amount of urban and second largest population in a tornado path. Supporting this growth was the second largest absolute increase in buildings with an additional 612 bringing the total to 1,316, the second largest amount for present-day. All this growth was at the expense of forest land cover with the loss of 3,044.2 acres, but only a 3.6% loss as this tornado path still contains 80,532.7 acres of forest.

Estimating present-day fatalities

Because of the multiple datasets and changes between time-periods involved, a lot of numerical results are provided in table format. One could analyze these results and assume what the impacts would be if the tornadoes occurred present-day. This study utilizes the results to estimate the number of present-day fatalities for all 21 tornadoes. The other studies discussed earlier do not take this approach. Specifically, a regression analysis was performed to test the relationship between each tornado’s number of fatalities and the six variables of Fujita rating, tornado length, tornado width, 1990 urban land cover, 1990 total population, and 1985 number of buildings. Except for tornado width, each variable had a positive correlation coefficient (r). Total population (p = 0.003), tornado length (p = 0.050) and Fujita rating (p = 0.027) were statistically significant at the p < 0.05 level. The variance inflation factors (VIF) were very high amongst all variables, ranging from 8 to 95. To remove the multicollinearity between these variables, a separate regression was conducted using only the three statistically significant variables of Fujita rating, tornado length, and total population (Table 9). Total population has the highest correlation coefficient (r) of 0.839 while tornado length has the weakest positive relationship (0.537). The multiple linear regression analysis yielded the following equations for each Fujita rating:

Table 9. Correlation coefficients and P-values. Model has a R2 value of 0.94 and adjusted R2 of 0.91.

		Population	Length	Fujita	F0	F1	F2	F3	F4	F5
Fatalities	r	0.839	0.537	0.634	-----	-----	-----	-----	-----	-----
	p < .01	0	0.003	0	0.145	0.579	0.55	0.432	0.006	0.001

• F0 FATALITIES = 0.37 + 0.007518(Population) – 0.1186(Length)

• F1 FATALITIES = -0.492 + 0.007518(Population) – 0.1186(Length)

• F2 FATALITIES = 1.254 + 0.007518(Population) – 0.1186(Length)

• F3 FATALITIES = 1.60 + 0.007518(Population) – 0.1186(Length)

• F4 FATALITIES = 5.64 + 0.007518(Population) – 0.1186(Length)

• F5 FATALITIES = -18.61 + 0.007518(Population) – 0.1186(Length)

The equations were tested by applying them to 1985 and present-day data generated by this study to estimate the number of fatalities for each time period. They predicted the exact number of 1985 fatalities but with slight differences amongst eleven of the twenty-one tornadoes (Table 10). Fatalities are underestimated for six tornadoes with the largest being three fewer fatalities for the F4 Tionesta tornado. Conversely, it overestimates fatalities for five tornadoes with the largest being three additional fatalities for another F4 tornado, Corry. When applied to present-day population data, the equations estimate 68 total fatalities, seven fewer than 1985. Seven have additional fatalities, nine have fewer, and five exhibit no change. The largest decrease is the F4 Atlantic tornado with eleven fewer fatalities due to a population decrease of 1,200, the second largest population loss of all tornadoes. With the largest population loss of 1,328, the F5 Wheatland tornado has ten fewer fatalities. Of the nine projected with a decrease in fatalities, six had a population loss within their paths and three had an increase. However, these increases were minimal at 4, 4, and 18.

Table 10. Estimation of tornado fatalities for 1985 and 2020 using Fujita regression equations.

No.	Tornado	Fujita	Length	Width	Pop 1985	Pop 2020	Actual fatal	Predict fatal 1985	Predict fatal 2020	Difference
										Actual − 1985	Actual − 2020
334	Albion	4	14	400	1,128	705	12	12	9	0	−3
336	Linesville	2	4	200	50	33	1	1	1	0	0
337	Atlantic	4	56	350	2,008	808	16	14	5	−2	−11
338	Saegertown	3	23	277	156	226	0	0	1	0	1
339	Corry	4	28	440	91	246	0	3	4	3	4
340	Dorset	2	10	220	22	65	0	0	1	0	1
341	Centerville	3	8	427	22	26	2	1	1	−1	−1
343	Wheatland	5	47	450	5,612	4,284	18	18	8	0	−10
344	Tionesta	4	29	800	200	173	7	4	3	−3	−4
345	Thompson Run	1	5	130	0	4	0	−1	−1	−1	−1
346	Lamont	2	19	300	11	29	0	−1	−1	−1	−1
349	Tidioute	3	17	800	48	84	0	0	0	0	0
350	Moshannon State Forest	4	69	3330	402	1,977	0	0	12	0	12
353	Big Bend	2	6	150	1	9	0	1	1	1	1
354	Emlenton	0	5	300	73	71	0	0	0	0	0
355	Kane	4	29	1000	479	171	4	6	3	2	−1
356	Big Beaver	3	39	243	1,755	1,428	9	10	8	1	−1
359	Watsontown	4	19	910	271	916	6	5	10	−1	4
360	Penn Run	0	6	27	2	1	0	0	0	0	0
362	Hollenback Township	1	11	530	447	536	0	2	2	2	2
363	Unknown	1	0	17	0	0	0	0	0	0	0
	TOTAL	—	Miles	Yards	12,778	11,791	75	75	68	0	−7

All seven tornadoes with a projected increase in fatalities also had population increases. The largest projected increase is the F4 Moshannon State Forest tornado, which had no fatalities during the 1985 tornado outbreak. It’s projected to have twelve present-day fatalities given it experienced the largest population increase (1,575) of all tornadoes. The F4 Corry tornado is another with no fatalities in 1985 but is projected to have four present-day fatalities. It has the third largest population increase of 155. With the second largest population increase (645), the F4 Watsontown tornado is projected to have four additional fatalities, bringing its present-day total to ten.

Discussion

The goal of this study was to analyze the impacts of the 21 tornadoes that affected Pennsylvania during the May 1985 outbreak across two time periods to determine the vulnerability of a repeat event based on changes in land cover, population, and building locations. The key findings from this analysis are:

• land cover within the tornado paths notably changed from forest and agriculture to urban (developed open space, developed low intensity, developed medium intensity, and developed high intensity).

• there was a 55/45 split between the number of tornadoes that gained and lost population but the overall total population within the study area noticeably decreased.

• age cohorts within the tornado paths shifted from young adults to middle-age/elderly (aged 50 and over). Essentially, the Baby Boomers generation was most affected in 1985 and would still be today as they have aged and not been replaced by younger Generation X and Generation Y/Millennials.

• the number of buildings in the tornado paths has greatly increased.

• the number of fatalities would be lower today than in 1985.

• when considering the entire study area as a whole, the May 1985 tornado paths demonstrate their bull’s eye has shrunk. However, geographic differences exist that specifically illustrate where the bull’s eye has shrunk, expanded, and shifted.

The growth and expansion of urban land cover into the tornado paths would suggest tornado vulnerability has increased since 1985. Previous research (Ashley and Strader 2016; Ashley et al. 2014; Rosencrants and Ashley 2015; Strader et al. 2018) has demonstrated that a spatially expanding built environment leads to greater hazard impacts and increased disaster potential. Most often this expansion is coupled with a growing population that places more people and their possessions in harm’s way (Rae and Stefkovich 2000; Wurman et al. 2007; Hall and Ashley 2008; Paulikas and Ashley 2011; Ashley et al. 2014; Rosencrants and Ashley 2015; Ashley and Strader 2016; Strader et al. 2017). Ashley et al. (2014) termed this the expanding bull’s-eye effect. However, this study found an increase in urban land cover within all tornado paths, but it was not always coupled with a growing population. Nine suffered population loss which suggests a shrinking and/or shifting bull’s eye effect. Fortunately, the loss was greatest in the Big-Three, thus greatly shrinking the size of their overall bull’s eye. Reason(s) for nine tornadoes having an inverse relationship between urban land cover and population growth are not understood based on the scope of this study, but several explanations can be inferred. If the population declines weren’t so large, one could argue this pattern exemplifies urban sprawl of low-density, large lot single-family dwellings like that found in the aforementioned studies. Perhaps this urban land was once populated but is now vacant for economic reasons. Western Pennsylvania’s population, where the majority of tornadoes occurred, began declining in the 1970s when the steel, coal, and manufacturing industries that were an economic foundation began to diminish (Forstall 1995). This population loss is supported by data at the county level. Of the 25 counties (including three in Ohio and one in New York) impacted by the May 1985 tornadoes, 19 had decreases in population from 1990 to 2020. The Big-Three, with the largest population losses, exemplify this most. The Wheatland tornado traversed two counties, beginning in Trumbull County, Ohio and ending in Mercer County, Pennsylvania. Each experienced population loss of 30,995 and 10,351. Mercer County also hosted the Atlantic tornado along with Venango County, Pennsylvania, which lost 8,921 of its population. The Albion tornado began in Ashtabula County, Ohio and ended in Erie County, Pennsylvania with each declining by 3,710 and 4,696 persons. However, population loss at the county level did not always produce a decreasing population within tornado paths as ten tornado paths experienced a population increase within decreasing counties. Perhaps this is a product of urban sprawl.

Eleven tornadoes demonstrated an expanded bull’s eye effect. Here, growth in urban land cover was coupled with an increasing population, most notably the Moshannon State Forest tornado. It exemplifies ideal characteristics of the expanding bull’s eye: urban expansion where areas that were once forest and large farmlands have transitioned to exurban development with increasing populations. The tornado had the largest absolute increases in urban and population, which expanded its bull’s eye from zero fatalities in 1985 to twelve for present-day. The tornado’s 69-mile path spans three Pennsylvania counites, each with a population increase: Clearfield (2,460), Center (34,386), and Clinton (268). It’s the only tornado path entirely located in a county or counties with increasing populations.

An increasing population within a tornado’s path did not directly correlate with a projected increase in fatalities. Of the eleven tornadoes with a population increase, six had a projected increase in fatalities and four a decrease. The latter raises the question of how to evaluate their bull’s eye as it relates to the findings of this study. The original study and definition provided by Ashley et al. (2014) does not consider a model that estimates fatalities based on population change. In this case, does a model projecting a decrease in fatalities based on a population increase constitute an expanding or shrinking bull’s eye effect? The original definition states that “…humans and their possessions….are enlarging as populations grow and spread…” Therefore, tornadoes with an increasing population but a projected decrease in fatalities demonstrate an expanding bull’s eye. Regardless, even tornadoes with a shrinking bull’s eye should not be ignored or deemed safer, as noted by Rosencrants and Ashley (2015). They found declining populations in Chicago and St. Louis but argue this does not cause those areas to be less exposed than others. The declines suggest that the size of their bull’s eye center has decreased in terms of population. Nonetheless this does not imply these areas are less vulnerable to hazards because of a possible heightened susceptibility and reduction in coping capacity associated with older populations living alone (Morrow 1999), decrepit housing conditions, and/or poverty (Klinenberg 2002). Thus, the nine tornadoes with a decreasing population shouldn’t be ignored. Seven have at least one estimated fatality for present-day and collectively they represent just over half (56%) of all estimated fatalities. Further, the age cohorts within these tornadoes are primarily late middle-aged and elderly. The percentage increase in elderly is concerning given they are less likely to receive tornado watches and warnings via social media or wireless emergency alerts (WEAs) (Walters et al. 2020), mitigation that has contributed to a decline in tornado fatalities as noted by Doswell, Moller, and Brooks (1999) and Brooks and Doswell (2002). They also note this includes several social, technical, and scientific advancements in modern forecasting, improved warning dissemination, spotter networks, hazards education and mitigation strategies, and construction practices. These advancements are particularly noteworthy as great strides have been made in each since 1985. This raises the question of what role they play in potentially reducing fatalities.

This tornado outbreak presents a formidable challenge for local policy makers and emergency managers to implement mitigation strategies. One of the worst tornado outbreaks occurred in a region not frequented by tornadoes, especially large ones. Strader et al. (2018) noted that even in tornado prone regions, improvements to structures and communities primarily occurred after devastating events like the 2011 Super Tornado Outbreak (southeast U.S.), 2011 Joplin, MO tornado, and the 2013 Moore, OK tornado. Except that Pennsylvania already had its devastating event. It just occurred a long time ago and no specific forms of mitigation were implemented afterwards. This research can suggest that policy makers and emergency managers advocate for and implement the standard mitigation practices of improving building codes, retrofitting existing structures to be more wind-resistant, tornado sirens, safe rooms, and storm shelters for mobile home parks. However, Doswell (2005) notes that like with all preventative measures, tornado mitigation is a combination of scientific, economic, psychological, and political issues. Pennsylvania’s is also cultural. It is possible all memories of the outbreak have since faded. Myths that tornadoes, or at least destructive ones, don’t occur in the state as the hills and mountains provide protection still exist amongst the locals. Thus, policy makers and emergency planners wishing to implement mitigation must confront the issues noted by Doswell (2005), but their biggest task will be convincing the population such measures are necessary. This research argues that at a minimum mitigation must come in the form of education. This includes both educating the population on the 1985 tornado outbreak and on tornado safety. The Executive Summary of the 1985 survey report produced by NOAA (1985) provides eleven key findings with recommendations, one which supports this argument. The first finding states that the overall response by the NWS in providing watches and warnings was rated very good and would be excellent if the rarity of such a broadscale outbreak is considered in the evaluation. Thus, the population received timely alerts, but did they know how to respond? The second finding notes that community awareness varied across the three states with a higher level of awareness and corrective actions taken in Ohio. This was related to an outreach program directed by the manager at the NWS office in Cleveland and awareness programs conducted in schools. The NWS recommended these awareness activities and preparedness campaigns be considered models for the nation. Policy makers and emergency planners must therefore be joined by weather officials and educators to bring about awareness. This research informed them how changes in land use patterns and population may influence tornado disaster potential both locally and regionally.

This study made a concerted effort to obtain and use geospatial datasets representative of the 1985 and present-day time periods to allow for direct comparison across each. Regardless, various assumptions had to be made. NLCD data for 1992 and 2019 was obtained from the MRLC. Relying on historical land cover data from 1992 presents various research challenges, especially when comparing it to more recent data. First, the data was not reflective of the time period before or during the tornado outbreak but rather seven years afterward. This research does not assume it represents 1985 but accepts it is the closest time period available from the same provider of the 2019 data. Ultimately it serves as a starting point for analyzing land cover changes within the 1985 tornado paths. Second, Stehman et al. (2003) estimated that overall accuracy of the 1992 data at the Anderson Level II classification is 46% in the Mid-Atlantic region, which includes Pennsylvania, and is 70% at the Anderson Level I. The 2019 NLCD has greater accuracy with Level II and Level I at 87% and 90% (Wickham et al. 2023). The latter classification method is utilized in this research, primarily due to the MRLC using different methodologies to create each NLCD that prevent comparison between each at Level II. Therefore, one cannot assume all land cover changes are strictly anthropogenic as consideration to the accuracy of the 1992 data are warranted. The most consistent data is the U.S. Census block groups for 1990 and 2020. However, like the land cover data, the 1990 data does not represent the study area during the tornado outbreak. This too is an unfortunate circumstance that prevents one from concluding how the outbreak influenced population change since the 1980 decennial census. Further, this research estimated populations according to the percentage of urban land cover within each tornado path. Results could be skewed given accuracy issues with the historical land cover data and the MRLC’s different methods in creating each dataset. Last, the 1985 buildings dataset was based on historical DRGs, with some created between 1954 and 1971. Most were updated with photo revisions between 1970 and 1990. Thus, the 1985 buildings dataset is representative of that 20-year period. The present-day buildings dataset is based on geocoded addresses and 2023 aerial photographs. One would expect an increase in buildings from 1985 to present-day, but the increases found in this research may be larger than expected given the time period and production of the DRGs. In the end, this research assumes the greatest accuracy of results is associated with present-day data and those based on the 1985 time period or changes since then are less accurate. Fortunately, this gives policy makers, emergency planners, weather officials, and educators an accurate portrayal of present-day land cover, population, and buildings within the 1985 tornado paths.

Future research on the 1985 tornado outbreak could use this study as a framework but rather analyze changes over ten-year increments: 1990 to 2000, 2000 to 2010, and 2010 to 2020. This may better illustrate points in time when changes in land cover and population occurred that demonstrate an expanding, shrinking, or shifting bull’s eye. Post 1992 land cover data would allow for comparison at the more detailed Anderson level II (developed – open space, low intensity, medium intensity, high intensity) to identify sparsely and highly populated residential areas. Future analyses could be expanded beyond individual tornado paths and be conducted over a much broader area to analyze multi-county level changes. Hypothetical tornado paths (Hall and Ashley 2008; Rosencrants and Ashley 2015; Strader et al. 2018) could then be utilized to analyze tornado vulnerability in those areas that demonstrate significant change. In addition to quantifying the number of buildings that are vulnerable, Wurman et al. 2007 noted that building types (single-and dual-family housing units, high-rise office and apartment buildings) can be used to determine the probability of destruction or surviving a potential tornado. Future research could also include social factors (race, density, wealth) according to Cutter, Boruff, and Shirley (2003) and Simmons and Sutter (2005) that influence tornado vulnerability. Race can be influential due to language and cultural factors that become a barrier to communication and cooperation between groups difficult. The density of the built environment has been shown to be a statistically significant factor in social vulnerability to disasters. Consequently, the larger number of people over a smaller area will likely lead to increased vulnerability in tornado events. Higher density areas are also found to have lower home values. The belief is that wealthier persons will have more money for safer housing and other preventative measures for hazards mitigation. Simmons and Sutter (2005) found income to be statistically significant determinant in tornado fatalities and injuries, signifying wealthier communities are better able to protect themselves. Whatever the focus of future research on tornado vulnerability in the Commonwealth of Pennsylvania, it’s long overdue. The last assessment of the state’s regional tornado trends and risks was compiled over two decades ago by current Penn State faculty member Dr. Jon Nese and former Penn State faculty member Dr. Greg Forbes (Nese and Forbes 1998). Prior to their research, then state-climatologist Paul Dailey Jr. conducted a trends and risk analysis for all Pennsylvania counties up to year 1970 (Dailey 1970). Much change has occurred since then. As previously stated, the eastern U.S. may be experiencing an increase in the number of tornado days and there are issues of climate change as it relates to severe weather (Hoogewind, Baldwin, and Trapp 2017). Planners and weather officials would greatly benefit from another state-wide tornado vulnerability study. Hopefully this study on the May 1985 tornado outbreak can serve as a revitalization before a similar event occurs again.

Conclusion

A total of 21 tornadoes impacted Pennsylvania during the May 1985 tornado outbreak, resulting in 75 fatalities and 815 injuries in the Commonwealth. No studies in academic literature have analyzed this outbreak. This research attempts to fill that void by analyzing the outbreak’s impacts on land cover, population, and locations of buildings between time-periods representative of 1985 and present-day. The goal was to analyze changes since 1985 to determine if tornado vulnerability changed for present-day if the event were to occur again and ultimately estimate the number of fatalities. From 1985 to present-day, the amount of urban land cover within the paths of all 1985 tornadoes greatly increased while the amount of forest and agriculture land greatly decreased. Similarly, the number of buildings increased substantially. Despite those increases, the overall total population within the paths of all tornadoes decreased across the time-period. However, this decrease was not uniform across all tornadoes with nine tornado paths having a population increase and thus exhibiting characteristics of the expanding bull’s eye effect. A regression analysis demonstrated the number of 1985 fatalities has a statistically significant relationship (p < 0.01) between Fujita rating, tornado length, and total population within the path of each tornado. A higher Fujita rating, longer tornado length, and higher total population were associated with a higher number of fatalities. The regression was applied to present-day data, and it was estimated that seven fewer fatalities would occur, primarily due to western Pennsylvania’s declining population. Results from this study can be used to provide further insight into the expanding bull’s eye effect, especially as it relates to rural areas at both a national and global scale. It could serve as a catalyst for future research to analyze Pennsylvania’s current tornado trends and risks, especially as it relates to climate change. The May 1985 tornado outbreak is a reminder that under the right atmospheric conditions, tornadoes and a tornado outbreak can occur in an uncommon region.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

Data analyzed in this study were obtained from a variety of sources. Some are a re-analysis of existing data and others were created by the author (available upon request). All are openly available at locations cited throughout the Data and Methods section and subsequently in the Reference section.