Severe Weather Events

Having analysed the data from the National Weather I have found that the heaviest damage, both in terms of injuries and fatalities as well as economic is caused by severe winds and floods. The single most costly storm is the Tornado, dwarfing all other storms.

Data Processing

fileURL <- "https://d396qusza40orc.cloudfront.net/repdata%2Fdata%2FStormData.csv.bz2"
destFile <- "StormData.csv.bz2"

## download.file(fileURL, destFile)

data <- read.csv(destFile, stringsAsFactors = FALSE)

I downloaded the StormData.csv.bz2 file and read it into a dataframe.

library(dplyr)

## 
## Attaching package: 'dplyr'
## 
## The following objects are masked from 'package:stats':
## 
##     filter, lag
## 
## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

library(ggplot2)

data <- select(data, c(EVTYPE, FATALITIES, INJURIES, PROPDMG, CROPDMG)) %>%
      mutate(FatalPlusInjuries = FATALITIES + INJURIES) %>%
      mutate(TotalDamage = PROPDMG + CROPDMG) %>%
      filter(FatalPlusInjuries >= 1 & TotalDamage >= 1)

data$EVTYPE <- as.factor(data$EVTYPE)

I then selected the Storm Type (EVTYPE), the health and economic damage, calculated the Fatalities plus Injuries (FatalPlusInjuries) and Total Damage (TotalDamage). I then removed events that had no damage.

summdata <- summarise(group_by(data, EVTYPE),
                      Injuries = sum(INJURIES),
                      Fatalities = sum(FATALITIES),
                      FatalPlusInjuries = sum(FatalPlusInjuries),
                      PropertyDamage = sum(PROPDMG),
                      CropDamage = sum(CROPDMG),
                      TotalDamage = sum(TotalDamage))

healthsummary <- select(summdata, c(EVTYPE, Injuries, Fatalities, FatalPlusInjuries)) %>%
            arrange(desc(FatalPlusInjuries)) %>%
            filter(FatalPlusInjuries > 1000)

damagesummary <- select(summdata, c(EVTYPE, PropertyDamage, CropDamage, TotalDamage)) %>%
      arrange(desc(TotalDamage)) %>%
      filter(TotalDamage > 10000)

I then summarised the data and created a summary for health and another for economic damage. I limited the health to storms where over 1,000 people were injured or died. I limited the economic damage to storms where over 10,000 in damage was wreaked.

Results

Here are the most damaging storm types.

Health (Death + Injuries)

healthsummary

## Source: local data frame [9 x 4]
## 
##              EVTYPE Injuries Fatalities FatalPlusInjuries
## 1           TORNADO    83283       5066             88349
## 2             FLOOD     6734        358              7092
## 3         TSTM WIND     2725        165              2890
## 4       FLASH FLOOD     1490        590              2080
## 5         ICE STORM     1842         40              1882
## 6 HURRICANE/TYPHOON     1275         62              1337
## 7 THUNDERSTORM WIND     1060         85              1145
## 8      WINTER STORM     1004         80              1084
## 9         HIGH WIND      877        151              1028

with(healthsummary, {
      qplot(FatalPlusInjuries, EVTYPE)
})

Economic Damage

damagesummary

## Source: local data frame [12 x 4]
## 
##                EVTYPE PropertyDamage CropDamage TotalDamage
## 1             TORNADO      877418.96   26402.09   903821.05
## 2           TSTM WIND      107854.43    3075.75   110930.18
## 3         FLASH FLOOD       61162.69    8306.20    69468.89
## 4           HIGH WIND       41247.50    1740.69    42988.19
## 5   THUNDERSTORM WIND       38986.20    1405.50    40391.70
## 6               FLOOD       29230.63   10369.85    39600.48
## 7  THUNDERSTORM WINDS       31769.80    1321.55    33091.35
## 8            WILDFIRE       19526.80    1068.20    20595.00
## 9           LIGHTNING       20311.00      13.50    20324.50
## 10       WINTER STORM       14963.91     293.00    15256.91
## 11               HAIL       10738.05    3463.00    14201.05
## 12         HEAVY SNOW       11034.74     170.00    11204.74

with(damagesummary, {
      qplot(TotalDamage, EVTYPE)
})

Combined Worst Storm Types

matrixsummary <- inner_join(healthsummary, damagesummary)

## Joining by: "EVTYPE"

with(matrixsummary, {
      qplot(FatalPlusInjuries, TotalDamage, color = EVTYPE)
})