Brandon Bartell
We analysed National Weather Service storm data recorded in the United States between 1950 and 2011 to assess what events have the most destructive impact to the economy and population health. After exploring the data, we elected to use property destruction value as a proxy for the economic impact of weather events and fatalities as a proxy for the impact to the population's health. After processing the data to explore the total impact each weather event had both socially and economically, we determined that tornados are by far the most destructive event by these standards. Tornados were the leading cause of fatalities and property damage among all weather events between 1950 and 2011, causing over $35 billion in property damage and over 4000 fatalities.
The data was loaded into R using read.csv()
#load dataset
setwd("C:/Users/Brandon/Desktop/Reproducible Research/CP2/")
library(R.utils)
bunzip2("repdata-data-StormData.csv.bz2")
data<-read.csv("repdata-data-StormData.csv")
#data<-read.csv("repdata-data-StormData.csv")
We used property damage value as a proxy for the economic consequences of a given event in the data. In the interest of getting a number for the property damage value, we selected only rows which had non-zero entries for PROPDMG. Then we used the designated exponent markers from the documentation (K, M, and B) to calculate the total value of the property damage in a column labeled DMGNUM. Finally, we created a data frame with only the two features of interest, event type and property damage value, then summed all of the property damage by event type. For ease of use, we sorted them in decreasing order of propert value damage.
#subset data by propert damage value and event type
nodmg<-which(data$PROPDMG!=0)
dmg<-data[nodmg,]
exp<-c("K","M","B")
goodexp<-which(dmg$PROPDMGEXP %in% exp)
dmg2<-dmg[goodexp,]
dmg2["DMGNUM"]<-NA
dmg2$DMGNUM[which(dmg2$PROPDMGEXP=="K")]=dmg2$PROPDMG[which(dmg2$PROPDMGEXP=="K")]*1e3
dmg2$DMGNUM[which(dmg2$PROPDMGEXP=="M")]=dmg2$PROPDMG[which(dmg2$PROPDMGEXP=="M")]*1e6
dmg2$DMGNUM[which(dmg2$PROPDMGEXP=="B")]=dmg2$PROPDMG[which(dmg2$PROPDMGEXP=="B")]*1e9
dfdmg<-data.frame(dmg2)
ndfdmg<-subset(dfdmg,select=c("EVTYPE","DMGNUM"))
dmgbyev<-aggregate(ndfdmg$DMGNUM,by=ndfdmg["EVTYPE"],FUN=sum)
dmgbyevord<-dmgbyev[order(-dmgbyev$x),]
We used fatalities as a proxy for harm to the population health. Consequently, we wanted to subset the data by event type and number of fatalities. To do this, we removed events which had no fatalities and then created a data frame with only the two features of interest, event type and fatalities. Finally we summed the fatalities by event type and sorted them in order of decreasing total fatalities.
#subset data by fatalities and event type
nofat<-which(data$FATALITIES!=0)
nofatdat<-data[nofat,]
fat<-data.frame(nofatdat)
fat2<-subset(fat,select=c("EVTYPE","FATALITIES"))
fatbyev<-aggregate(fat2$FATALITIES,by=fat2["EVTYPE"],FUN=sum)
fatbyevord<-fatbyev[order(-fatbyev$x),]
names(fatbyevord)<-c("Event_Type","Total_Fatalities")
Now that we have the total property damage value and fatalities, we can plot the top 5 most destructive events in each category.
names(fatbyevord)<-c("Event_Type","Total_Fatalities")
names(dmgbyevord)<-c("Event_Type","Total_Property_Damage_Value")
barplot(fatbyevord[1:5,2], main="Fatalities by Event Type", ylab="Fatalities",
names.arg=fatbyevord[1:5,"Event_Type"],las=1,cex.names=0.75)
We can see that tornados are far and away the most deadly event, causing more than twice as many deaths, 5633, than the next 4 most deadly events combined, 4634.
par(mar=c(5.5,5,4.1,2.1))
barplot(dmgbyevord[1:5,2], main="Property Damage by Event Type",
ylab="Property Damage Value ($)",names.arg=dmgbyevord[1:5,"Event_Type"]
,las=1,cex.names=0.75, yaxt="n")
axis(4)
Tornados also appear to cause the most economic damage by property loss, $35.47 billion in the time frame covered by this data set (1950-2011).
Interestingly, tornados are the only event in the top 5 most destructive in terms of property damage AND human lives.