A Dasymetric Method to Estimate Tornado Casualties Spatially

Poster presented at the 2016 Severe Convective Storms Workshop
Columbia University, New York City, NY

Transcript

1. ADasymetricMethodtoEstimateTornadoCasualtiesSpatially Tyler Fricker (tfricker@fsu.edu, @TylerFricker), James B. Elsner, Victor Mesev & Thomas H. Jagger Department of Geography. Florida State University Why Did We Do This? Ashley (2007) provides a spatial and temporal analysis of tornado fatal- ities using location information provided by Grazulis (1990) and Storm Data . The locations are by either the nearest town (if available) or the nearest county seat (when only the county where the fatality occurred is known). Using these data over the period 1950 to 2004 and group- ing by 60 km cells they map tornado deaths across the United States, although locational accuracy may be variable, most notably of data col- lected before 1985, where the location of some fatalities is incorrectly coded in Storm Data. In response, we suggest a technique that allocates the total number of casualties reported for each separate tornado in proportion to the un- derlying population geography that contains its path. The technique uses dasymetric mapping principles in that volumetric data (total ca- sualties and total population) are spatially reapportioned across areal data (buffered tornado tracks and census zones) respectively. Dasymet- ric calculations for an individual tornado are then summed across all tornadoes to create maps that represent casualties. Where Do People Live? 2010 Population Density (persons per km2) 32°N 34°N 36°N 38°N 40°N 100°W 95°W 90°W 85°W 0 0 1 4 16 64 256 1024 What Do the Data Look Like? 0 100 200 300 400 500 1955 1965 1975 1985 1995 2005 2015 Year Number of Tornado Deaths A 0 10 20 30 40 50 1955 1965 1975 1985 1995 2005 2015 Year Number of Killer Tornadoes B 0 500 1000 1500 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Number of Tornado Deaths C 1 10 100 1 10 Number of Deaths Number of Tornadoes D How Do We Do It? We apply a dasymetric method that spatially integrates tornado paths with the underlying population. Our central premise is that the like- lihood of casualties increases the greater the population, therefore we assign to each grid cell within the tornado path a fraction of the total number of casualties associated with the tornado. This fraction is based on the area of the tornado path that falls within the cell and the cell’s population density. The fractional number of casualties are summed over each tornado-casualty event affecting the cell over the study pe- riod to get an estimated number of casualties. 0 0 0 1 2 0 1 91 1 0 3 1 0 0 0 Population Density (persons per square km) < 5 5−50 51−100 101−1000 > 1000 Maps of Tornado Casualties Estimated Number of Tornado Deaths 32°N 34°N 36°N 38°N 40°N 100°W 95°W 90°W 85°W 1 3 10 32 100 316 Estimated Number of Tornado Injuries 32°N 34°N 36°N 38°N 40°N 100°W 95°W 90°W 85°W 1 3 10 32 100 316 1000 3162 Clusters of Higher Casualties Local Moran's I [Deaths] 32°N 34°N 36°N 38°N 40°N 100°W 95°W 90°W 85°W −1 −0.75 −0.5 −0.25 0 0.25 0.5 0.75 >1 Local Moran's I [Injuries] 32°N 34°N 36°N 38°N 40°N 100°W 95°W 90°W 85°W −1 −0.75 −0.5 −0.25 0 0.25 0.5 0.75 >1 How Good is the Method? A Estimated Number of Tornado Deaths (1955−2014) 0 4 8 12 16 20 B Actual Number of Tornado Deaths (1955−2014) 0 4 8 12 16 20 How Good is the Method? 0 5 10 15 20 0 5 10 15 20 Actual Number of Deaths Estimated Number of Deaths A 0 100 200 300 400 0 100 200 300 400 Actual Number of Injuries Estimated Number of Injuries B Conclusions Casualty statistics are well documented at the level of individual tor- nadoes, but less is known at a more local level, especially along the tornado path. We outline a methodology that applies a dasymetric principle to link tornado paths with tornado-induced casualty data and disaggregated population to create maps of estimated numbers of casu- alties, both deaths and injuries. Results show a relatively uniform dis- tribution of across a large area covering “Tornado Alley” with slightly higher rates in the mid South, particularly in northern Mississippi and Alabama, and in many metropolitan areas. The method is validated using county-level casualty data (here shown as the state of Tennessee). Better maps can be achieved with better data. For instance, population density at the year the tornado occurred would likely improve the spatial depiction of casualties especially for torna- does with paths across areas where population gradients have changed dramatically. Additionally not all tornadoes create a straight-line dam- age path. Greater detail on the path polygon would improve the results, especially if the method is applied to a grid with small cells.