Capturing Damage Assessment with a Spatial Video: An Example of a Building and Street-Scale Analysis of Tornado-Related Mortality in Joplin, Missouri, 2011

Abstract

To reduce the potential loss of life during future significant tornadoes, it is important to understand building and street-scale patterns of damage and mortality from past events. To achieve this, methods of posttornado data collection and damage classification should be rethought. This article addresses these needs through the use of a spatial video and a new Tornado Injury Scale, which has evolved from the Enhanced Fujita Scale. Building damage for Joplin, Missouri, collected three weeks after the May 2011 tornado, is analyzed using a filtering approach more commonly employed in epidemiological studies. Tornado-related fatalities are linked to damage rate maps constructed at two relatively fine scales. Mortality patterns are also investigated with regard to social vulnerability, especially the disproportionate burden faced by the elderly. Results reveal fine-scale damage patterns even within the main tornado path, often with variations occurring along the same street segment. In Joplin the number of elderly fatalities was significantly higher than for other cohorts. This finding is consistent within neighborhoods and not just as a result of single-site losses, as has been found in other disasters. Spatial video footage is also reinvestigated after the analysis for damage trends by building type and physical setting, with a summary of these observations provided in the discussion. This article is a contribution to understanding the fine spatial scale variation of built environment fragility within a tornado path.

为了减低未来重大龙捲风可能造成的生命损失, 从过往的事件中了解建筑与街道尺度的损害模式及死亡率便相当重要。为了达到上述目的, 我们应该重新考量龙捲风过后的资料搜集与损害分类方法。本文透过运用空间录像, 以及由改良的籐田级数引伸出的新龙捲风伤害级数, 因应上述的需求。本文运用在流行疾病学研究中更广为运用的过滤方法, 分析 2011 年密苏里州乔普林市遭遇龙捲风袭击三週后所搜集到的建物损害资料。与龙捲风相关的伤亡, 连结至在两个相对精细的尺度中建构的损害率地图。本研究同时调研死亡模式与社会脆弱性的关联, 特别是老年人所面临的不成比例的重担。研究结果揭露了精细尺度的损害模式, 即便是在主要的龙捲风路径当中, 亦经常沿着同一条街道产生不同的变异。在乔普林市, 老年人的死亡人数远较其他年龄群体来得高。此一发现在各邻里中皆相同, 而并非如同其他灾害中的发现一般, 仅是某单一地点的损失。本文依建物类型与建成环境分析灾害倾向后, 亦重新探讨空间录像片段, 并在讨论中提供这些观察的概要。本文对于理解龙捲风路径中建成环境脆弱性在细微空间尺度的变异做出贡献。

Para reducir la potencial pérdida de vidas humanas durante futuros tornados peligrosos, es importante entender los patrones de daño y mortalidad en eventos pasados a escala de edificios y de calles. Para lograr esto es necesario volver a los métodos de recolección de datos con posterioridad al tornado y la clasificación de daños. Este artículo aborda estas necesidades mediante la utilización de un video espacial y una nueva Escala de Lesiones por Tornado, una versión derivada de la Escala Fujita Mejorada. Los datos de daños a los edificios de Joplin, Missouri, recogidos tres semanas después del tornado de mayo de 2011, se analizaron mediante un enfoque de filtrado, más comúnmente utilizado en estudios epidemiológicos. Las fatalidades causadas por tornados son relacionadas con mapas de tasas de daño, construidos a dos escalas relativamente finas. También se investigaron los patrones de mortalidad, en consideración a la vulnerabilidad social, especialmente en lo que concierne al impacto desproporcionado que deben enfrentar los viejos. Los resultados ponen de manifiesto patrones de daño a escala fina incluso dentro de la ruta principal del tornado, a menudo con variaciones a lo largo del mismo segmento de una calle. El número de fatalidades de adultos mayores en Joplin fue significativamente más alto que el de otras cohortes. Este hallazgo es consistente dentro de los barrios y no solo como de resultado de pérdidas en un sitio particular, como se ha establecido en otro tipo de desastres. La grabación espacial en vídeo también fue re-investigada después del análisis de las tendencias de daños por tipo de edificio y escenario físico, con un resumen de estas observaciones exhibidas en la discusión. Este artículo contribuye en el entendimiento de la variación de la escala espacial fina de la fragilidad del ambiente construido situado en la ruta de un tornado.

Key Words:

• damage assessment
• elderly
• health vulnerability
• spatial video
• tornado.

关键词:

• 损害评估
• 老年人
• 健康脆弱性
• 空间录像
• 龙捲风

Palabras clave:

• evaluación del daño
• adultos mayores
• vulnerabilidad de la salud
• vídeo espacial
• tornado

Acknowledgments

Both authors would like to thank Jacqueline W. Curtis at Kent State University, Louisa Holmes at the University of Southern California, and Marc Levitan at the National Institute of Standards and Technology for their helpful comments regarding this article. Both authors would also like to thank the article reviewers who provided invaluable insights.

Notes

1. A synthesized summary of the warning sequence was a watch issued at 1:30 p.m. and a warning for the general vicinity given at 5:09 p.m., followed by a three-minute siren at 5:11 p.m. for Jasper and Joplin Counties. The first touchdown southwest of Joplin City occurred at 5:34 p.m., with a second three-minute siren at 5:38 p.m. for both counties. At approximately the same time, the beginning of EF4 damage began to occur (NWS 2011).

2. For purposes of this article, these are both simply referred to as apartments.

3. The tornado was classed an F4, killed forty-seven people, and destroyed or made uninhabitable 3,000 (out of 4,800) homes.

4. This leads to additional questions, such as how spatial vulnerability should be measured and analyzed. For example Ebert, Kerle, and Stein (2009) used proxy variables derived from the contextual object-oriented analysis of remotely sensed data to calculate social vulnerability assessment.

5. Phillips and Morrow's (2007) article on social vulnerability uses Mileti's seven stages of processing a warning (Mileti 1999), which can be paraphrased as receiving, believing, confirming, and assessing (risk to self) and then three steps of action, including shall we do, can we do, and what to do. It is the last three steps of action where the greatest vulnerability lies for the elderly.

6. For example, Glass et al. (1980) adopted American Red Cross damage assessment coding ranging from 0 (no damage) to 4 (irreparable). Access to street sheet information collected by the American Red Cross is generally hard to acquire because of confidentiality issues, a lack of digital form, and a general institutional loss in knowing exactly where data from past events reside. Further problems with this windshield survey–style data collection by volunteer teams is a lack of verification of the scores. Comstock and Mallonee (2005) used a scale including no damage, some walls damaged, and completely collapsed. In other studies, data collected by the researcher have involved walking the streets or using mail, community, and telephone surveys (see Brown et al. 1999).

7. Examples of this type of oblique photography are available at http://www.pictometry.com.

8. For the full series of tables see the NOAA web page (http://nationaloceanicandatmosphericadministration) at http://www.spc.noaa.gov/faq/tornado/ef-scale.html.

9. One of the criticisms and recommendations made by Doswell, Brooks, and Dotzek (2009) was a lack of consistency, or “experience,” as the same field survey team will not attend every tornado event. One partial solution was greater attention paid to metadata capture, but the spatial video approach described in this article helps solve this issue by allowing for secondary expert analysis as a permanent visual record accompanies the field survey.

10. Initial coding tests were used to establish the effectiveness of TIS score levels. Modifications occurred with level 5 as more information was added to fill the gap between 4 (loss of the majority of roof material) and 6 (removal of the roof). In many instances, TIS scores of 10 have conservatively been classed as 9 because some debris is still present. This is not important, as our objective is to assess risk to life and there is little difference between the two scores. Given the previously described limitations with aerial imagery, it is likely that the proposed TIS score is best suited for ground-based surveys. The ImageCat team providing damage assessment for the 2011 tornadoes described using a similar version of the EF scale, although maps presented in their after-action report only used FEMA's simplified damage classification scheme.

11. The postevent imagery was three-inch aerial photography flown on 23 May 2011 (see http://msdisimagery-tc.col.missouri.edu/ArcGIS/services).

12. Alternatives to digitizing for future events might include preexisting building footprint or property parcel GIS layers. Of these, parcels are more commonly available for most urban areas but lack the detail of building size and placement.

13. Apartments were included in this category.

14. DMAP software requires distances of miles. Figures in this article include scales drawn in kilometers in keeping with usual scientific convention.

15. The original data source used by the U.S. Army Corps of Engineers was a point shapefile of building damage provided by the National Geospatial Intelligence Agency, using FEMA tornado damage classification (Catastrophic, Extensive, Moderate, Limited). To create the raster a numeric value was assigned to each categorical class (4 = Catastrophic, 3 = Extensive, 2 = Moderate, 1 = Limited), then an inverse distance weighted interpolation was performed (Michael Dulin, Army Corps of Engineers cartographer, personal communication). For Figure 5 a reverse coloring scheme (dark to light) is used for the interpolation to ease comparison.

16. Fatalities that were not included were those that occurred in St. John's Hospital, during search and rescue, or had questionable location or cause of death.

17. A fine-scale analysis was performed whereby each residential death was overlaid onto the 2010 census block population. Rates were then calculated for each block in terms of the proportion of elderly deaths and the proportion of elderly residing in each block. This type of fine-scale investigation is vulnerable to a small numbers effect; however, both blocks with the highest number of deaths (four) had 75 percent and 25 percent of the deaths being among the elderly, with the associated proportion of elderly population in each block being 25 and 4 percent, respectively. Of the eight blocks where three deaths were recorded, the proportion of deaths among the elderly exceeded the elderly population proportion in five situations, with one block having no elderly population residing within, and therefore a zero count of elderly deaths was expected. These results suggest that there was a disproportionate impact to the elderly even at this finest of census aggregations.

18. This did not really mean one fifth of all buildings as some residences had multiple fatalities.

19. Statistically different at p = 0.01 using a difference of proportions t test.

20. A future analysis could determine to what degree buildings were truly brick built as compared to brick facades.

21. In the example given by Speheger and Smith (2006), the path of a 1999 tornado struck a community eight miles outside of the warning area because of radar imprecision and an erroneous linear extrapolation.