Analysis of the Most Dangerous and Economically Damaging Storm Event Types

This report analyzes different natural storm event types by the danger they pose to humans and the economic damage to property and crops. The danger to humans is measured by recorded injuries and fatalities, while the damage to property and crops is measured in US dollars. The data come from the US National Oceanic and Atmospheric Administration’s (NOAA) storm database and was recorded from 1950 to 2011. Tornadoes are the most harmful storm event type in terms of injuries and fatalities while floods cause the most damage to property and crops.

Data Processing

storm_data <- read.csv("repdata_data_StormData.csv.bz2")
# Combine TSTM wind, thunderstorm wind and thunderstorm winds
storm_data$EVTYPE[storm_data$EVTYPE=="TSTM WIND"] <- "THUNDERSTORM WIND"
storm_data$EVTYPE[storm_data$EVTYPE=="THUNDERSTORM WINDS"] <- "THUNDERSTORM WIND"

Results

Most Harmful Storm Event Types Related to Public Health

library(dplyr)
total_fatalities_injuries <- group_by(storm_data, EVTYPE) %>%
  summarise(total_injuries=sum(INJURIES), total_fatalities=sum(FATALITIES)) %>%
  arrange(-total_injuries-total_fatalities)
library(reshape2)
total_fatalities_injuries <- melt(total_fatalities_injuries[1:10,], id=c("EVTYPE"))
library(ggplot2)
ggplot(total_fatalities_injuries, aes(x=reorder(EVTYPE, -value),y=value,fill=variable)) +
  geom_bar(stat="identity") +
  ggtitle("Top Ten Event Types by Injuries and Fatalities") +
  xlab("Event Type") +
  ylab("Number of Cases") +
  theme(axis.text.x = element_text(angle=90)) +
  scale_fill_discrete(labels=c("Total Injuries", "Total Fatalities"))

Over the time period 1950-2011, tornadoes have caused the highest number of recorded injuries and fatalities in the United States. Municipalities should prioritize protecting against tornadoes by building more tornado shelters and tornado resistant structures.

Most Harmful Storm Event Types Related to Economic Damage

The chunk below transforms the values in PROPDMG and CROPDMG to actual dollar values using the exponents contained in PROPDMGEXP and CROPDMGEXP.

convertMultiplier <- function(mult) {
  mult <- as.character(mult)
  if (mult == "B" || mult == "b") {
    return(10^9)
  } else if (mult == "M" || mult == "m") {
    return(10^6)
  } else if (mult == "K" || mult == "k") {
    return(10^3)
  } else if (mult == "H" || mult == "h") {
    return(10^2)
  } else if (!is.na(suppressWarnings(as.numeric(mult)))) {
    num <- as.numeric(mult)
    stopifnot(num >= 0 && num < 10)
    return(10^num)
  } else {
    stopifnot(mult == "" ||
                mult == " " ||
                mult == "-" ||
                mult == "?" ||
                mult == "+")
    return(1)
  }
}
storm_data$prop_multiplier <- sapply(storm_data$PROPDMGEXP, convertMultiplier)
storm_data$PROPDMG <- storm_data$prop_multiplier * storm_data$PROPDMG
storm_data$crop_multiplier <- sapply(storm_data$CROPDMGEXP, convertMultiplier)
storm_data$CROPDMG <- storm_data$crop_multiplier * storm_data$CROPDMG

total_property_crop_damage <- group_by(storm_data, EVTYPE) %>%
  summarise(total_property_damage=sum(PROPDMG), total_crop_damage=sum(CROPDMG)) %>%
  arrange(-total_property_damage-total_crop_damage)
total_property_crop_damage <- melt(total_property_crop_damage[1:10,], id=c("EVTYPE"))
ggplot(total_property_crop_damage, aes(x=reorder(EVTYPE,-value),y=value/10^6,fill=variable)) +
  geom_bar(stat="identity") +
  ggtitle("Top Ten Event Types by Property and Crop Damage") +
  xlab("Event Type") +
  ylab("Damage in Millions of US Dollars") +
  theme(axis.text.x = element_text(angle=90)) +
  scale_fill_discrete(labels=c("Total Property Damage", "Total Crop Damage"))

Over the time period 1950-2011, floods have caused the highest dollar value of property damage in the United States. Municipalities should prioritize building more flood resistant structures such as levees to protect homes from flood damage. Droughts have caused the most crop damage.