The Most Harmful and Destructive Severe Weather Events in the USA (1950-2011)

This document analyzes the most destructive and harmful severe weather events in the USA, using the National Oceanic and Atmospheric Administration’s (NOAA) storm database that tracks characteristics of major storms and weather events and estimates of fatalities, injuries, and property and crop damage between 1950 and 2011. The below analysis charts the top five severe weather events that produced the most fatalities and injuries and caused the most destruction to property and crop (measure in USD). Five variables were mainly used. EVTYPE capture the type of severe weather event. FATALITIES and INJURIES show the number of people killed and injured, respectively, during a weather event. PROPDMG and CROPDMG show the cost of property and crop damage, respectively, during a weather event. Cross tabulation and bar polots were used in the statistical analysis.

Data Processing

The NOAA data were download manually from the Course Web site and moved into the working directory (Note: This could have been automated, but slow internet in my country means I have to wait two hours to just download this file, even if cached). The command read.csv is used to read the .bz2 file without the need of unzipping. Cache was activated to save time reading the dataset.

data <- read.csv("repdata-data-StormData.csv.bz2")

Results

The following creates a cross tabulation between the weather events and number of fatalities and injuries, respectively. It then sorts the data and saves the highest five numbers of fatatlities and injuries into other variables which are plotted below.

Fatalities_crosstab <- xtabs(FATALITIES ~ EVTYPE, data=data)
a <- abs(head(sort(-Fatalities_crosstab),5))
Injuries_crosstab <- xtabs(INJURIES ~ EVTYPE, data=data)
b <- abs(head(sort(-Injuries_crosstab),5))

The following bar plots show that the top five severe weather events in terms of fatalities are tornados, excessive heat, flash floods, heat and lightning, and the top five severe weather events in terms of injuries are tornados, TSTM wind, excessive heat, and lightening.

par(mfcol=c(1,2), mar=c(9, 5, 5, 2))
barplot(a, las=3, ylim=c(0,6000), col=rainbow(20), ylab="Number of Fatalities", main="Top Five Weather Events \n by number of fatalities")
barplot(b, las=3, ylim=c(0,90000), col=rainbow(20),  ylab="Number of Injuries", main="Top Five Weather Events \n by number of injuries")

The following creates a cross tabulation between the weather events and cost incured from property and crop damage, respectively. It then sorts the data and saves the highest five costs of property and crop damage into other variables which are plotted below.

Property_Damage_crosstab <- xtabs(PROPDMG ~ EVTYPE, data=data)
c <- abs(head(sort(-Property_Damage_crosstab),5))
Crop_Damage_crosstab <- xtabs(CROPDMG ~ EVTYPE, data=data)
d <- abs(head(sort(-Crop_Damage_crosstab),5))

The following bar plots show that the top five severe weather events in terms of cost incurred from property damage are tornados, flash floods, TSTM winds, floods, and Thunderstorm wind, and the top five severe weather events in terms of cost incurred from crop damage are hail, flash floods, thunderstorm winds, and tornados.

par(mfcol=c(1,2), mar=c(11, 5, 5, 2))
barplot(c, las=3, col=rainbow(20), ylab="Totol Cost of Damage in USD", main="Top Five Weather Events \n by property damage cost")
barplot(d, las=3, col=rainbow(20),  ylab="Totol Cost of Damage in USD", main="Top Five Weather Events \n by by property damage cost")

It seems that roughly the same types of severe weather events tend to cause the most fatalities and injuries and contribute to the highest cost in property and crop damage, especially tornados, flash floods and TSTM winds. It is recommended that high priority is given to researching such weather events.