APPLICATION OF MACHINE LEARNING TO ESTIMATION OF DEATHS, INJURY AND PROPERTY DAMAGE FROM TORNADOS IN DIFFERENT COMMUNITIES

Among the factors affecting urban community resilience to tornados are socio-economic status, severity and warning systems. Urban areas are usually a mosaic of areas with contrasting Socio-Economic Status (SES), ranging from prosperous, affluent communities with ready access to basements and shelters and timely and effective warning systems, to distressed communities, living in poverty, who live in trailers or low-quality housing stock, without basements or ready access to shelters and sometimes without timely warning systems. An F-3 tornado hit the relatively prosperous and affluent community of Andover, Kansas, on April 29^th at 8 PM, remaining on the ground for 21 minutes, but there were no deaths and few injuries. If the same tornado had occurred just 10 miles to the West, the results could have been quite different. The path of such a tornado would cross trailer parks and low-quality, outer suburban housing, as well as aging, low quality, inner suburban housing. Machine learning as the potential to use SES data and tornado strength and paths to estimate the rate of death, injuries and property damage.