Examining Population and Economic Damages of Severe Weather in the United States from 1950-2011
Synopsis:
This report explores the outcomes of the most impactful severe weather events with respect to population health and economic effects in the United States. The report examines over 90,000 events of severe weather from the NOAA Storm Database from years 1950 until 2011. The analyses conclude that tornados cause the highest number of both fatalities and injuries - in total 85% of the fatalities of the top seven fatality-causing events. However, looking at average injuries per event, blizzards and floods cause significant injuries. When analyzing economic effects-which the data separate into Property Damage and Crop Damage-the report concludes that Property Damage is much more likely to result from severe weather than Crop Damage. Damage from water events (storms and flooding) account for the most amount of property damage. Drought (lack of water) accounts for the highest amount of crop damage.
Data Processing for First Analysis:
After loading several packages,we’ll use the bzfile command to read in the entire csv file into R.
options(scipen=999)
options(warn=-1)
library(ggplot2)
library(gridExtra)## Loading required package: gridDat <- read.csv(bzfile("repdata-Data-StormData.csv.bz2"))We’ll peform the str() funtion to get a sense of what the data look like:
str(Dat)## 'data.frame': 902297 obs. of 37 variables:
## $ STATE__ : num 1 1 1 1 1 1 1 1 1 1 ...
## $ BGN_DATE : Factor w/ 16335 levels "1/1/1966 0:00:00",..: 6523 6523 4242 11116 2224 2224 2260 383 3980 3980 ...
## $ BGN_TIME : Factor w/ 3608 levels "00:00:00 AM",..: 272 287 2705 1683 2584 3186 242 1683 3186 3186 ...
## $ TIME_ZONE : Factor w/ 22 levels "ADT","AKS","AST",..: 7 7 7 7 7 7 7 7 7 7 ...
## $ COUNTY : num 97 3 57 89 43 77 9 123 125 57 ...
## $ COUNTYNAME: Factor w/ 29601 levels "","5NM E OF MACKINAC BRIDGE TO PRESQUE ISLE LT MI",..: 13513 1873 4598 10592 4372 10094 1973 23873 24418 4598 ...
## $ STATE : Factor w/ 72 levels "AK","AL","AM",..: 2 2 2 2 2 2 2 2 2 2 ...
## $ EVTYPE : Factor w/ 985 levels " HIGH SURF ADVISORY",..: 834 834 834 834 834 834 834 834 834 834 ...
## $ BGN_RANGE : num 0 0 0 0 0 0 0 0 0 0 ...
## $ BGN_AZI : Factor w/ 35 levels ""," N"," NW",..: 1 1 1 1 1 1 1 1 1 1 ...
## $ BGN_LOCATI: Factor w/ 54429 levels "","- 1 N Albion",..: 1 1 1 1 1 1 1 1 1 1 ...
## $ END_DATE : Factor w/ 6663 levels "","1/1/1993 0:00:00",..: 1 1 1 1 1 1 1 1 1 1 ...
## $ END_TIME : Factor w/ 3647 levels ""," 0900CST",..: 1 1 1 1 1 1 1 1 1 1 ...
## $ COUNTY_END: num 0 0 0 0 0 0 0 0 0 0 ...
## $ COUNTYENDN: logi NA NA NA NA NA NA ...
## $ END_RANGE : num 0 0 0 0 0 0 0 0 0 0 ...
## $ END_AZI : Factor w/ 24 levels "","E","ENE","ESE",..: 1 1 1 1 1 1 1 1 1 1 ...
## $ END_LOCATI: Factor w/ 34506 levels "","- .5 NNW",..: 1 1 1 1 1 1 1 1 1 1 ...
## $ LENGTH : num 14 2 0.1 0 0 1.5 1.5 0 3.3 2.3 ...
## $ WIDTH : num 100 150 123 100 150 177 33 33 100 100 ...
## $ F : int 3 2 2 2 2 2 2 1 3 3 ...
## $ MAG : num 0 0 0 0 0 0 0 0 0 0 ...
## $ FATALITIES: num 0 0 0 0 0 0 0 0 1 0 ...
## $ INJURIES : num 15 0 2 2 2 6 1 0 14 0 ...
## $ PROPDMG : num 25 2.5 25 2.5 2.5 2.5 2.5 2.5 25 25 ...
## $ PROPDMGEXP: Factor w/ 19 levels "","-","?","+",..: 17 17 17 17 17 17 17 17 17 17 ...
## $ CROPDMG : num 0 0 0 0 0 0 0 0 0 0 ...
## $ CROPDMGEXP: Factor w/ 9 levels "","?","0","2",..: 1 1 1 1 1 1 1 1 1 1 ...
## $ WFO : Factor w/ 542 levels ""," CI","$AC",..: 1 1 1 1 1 1 1 1 1 1 ...
## $ STATEOFFIC: Factor w/ 250 levels "","ALABAMA, Central",..: 1 1 1 1 1 1 1 1 1 1 ...
## $ ZONENAMES : Factor w/ 25112 levels ""," "| __truncated__,..: 1 1 1 1 1 1 1 1 1 1 ...
## $ LATITUDE : num 3040 3042 3340 3458 3412 ...
## $ LONGITUDE : num 8812 8755 8742 8626 8642 ...
## $ LATITUDE_E: num 3051 0 0 0 0 ...
## $ LONGITUDE_: num 8806 0 0 0 0 ...
## $ REMARKS : Factor w/ 436781 levels "","-2 at Deer Park\n",..: 1 1 1 1 1 1 1 1 1 1 ...
## $ REFNUM : num 1 2 3 4 5 6 7 8 9 10 ...For our first set of analyses, we’re going to be most interested in three variables: EVTYPE, FATALITIES, and INJURIES. We want to understand which categories of disasters have the most detrimental outcomes to public health.
Let’s subset our raw dataset to get rid of observations that had both zero injuries and zero fatalities. This data frame will be called Sub. We’ll then make two data frames - HiFatal and HiInjury and see what types of EVTYPE are responsible for the highest deaths and injuries.
Sub <- Dat[Dat$FATALITIES != 0,]
Sub <- Sub[Sub$INJURIES != 0,]We’ve now significantly cut down the number of rows in our dataset. Let’s do a summary() of the INJURIES and FATALITIES columns to see what we’re looking at.
summary(Sub$FATALITIES)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 1.000 1.000 1.000 2.862 2.000 158.000summary(Sub$INJURIES)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 1.00 2.00 5.00 29.82 20.00 1700.00For deaths, it appears the values are generally low (median is 1) with a very long tail. Similarly for injuries, the 75th percentile of most injuring events is 20 injuries. To get at the most damaging events, let’s continue to subset this dataset, but now we’ll create two sets - hifatal and hiinjury where we will only look at the 90th percentile and above for fatalities and injuries.
hifatal <- subset(Sub,Sub$FATALITIES >= (quantile(Sub$FATALITIES,.9)))
hiinjury <- subset(Sub,Sub$INJURIES >= (quantile(Sub$INJURIES,.9)))Let’s now use ddply to summarize the event types with the highest of fatalities or injuries to see which event types stand out.
library(plyr)
fatalsum <- ddply(hifatal,.(EVTYPE),summarise,total=sum(FATALITIES),mean=mean(FATALITIES))
fatalsum <- arrange(fatalsum,desc(total))
injurysum <- ddply(hiinjury,.(EVTYPE),summarise,total=sum(INJURIES),mean=mean(INJURIES))
injurysum <- arrange(injurysum,desc(total))Let’s look at the head() of our new data tables to get a snapshot of the highest death and injury-causing disasters, ordered by highest total deaths/injuries first and descending.
head(fatalsum,7)## EVTYPE total mean
## 1 TORNADO 3429 12.89098
## 2 EXCESSIVE HEAT 314 20.93333
## 3 FLASH FLOOD 66 11.00000
## 4 HEAT 54 10.80000
## 5 FLOOD 42 8.40000
## 6 WILDFIRE 34 11.33333
## 7 TSUNAMI 32 32.00000head(injurysum,7)## EVTYPE total mean
## 1 TORNADO 41556 196.9479
## 2 EXCESSIVE HEAT 4011 174.3913
## 3 FLOOD 2492 498.4000
## 4 ICE STORM 1568 1568.0000
## 5 HEAT 1320 165.0000
## 6 HURRICANE/TYPHOON 1200 400.0000
## 7 BLIZZARD 535 267.5000Data Processing for Second Analysis:
Our second analysis has to deal with tabulating total Property and Crop damages to see which event types have the most weakening economic effects. The raw data frame has two columns for both property and crop damages, DMG and DMGEXP. In the code book, we learn that the DMG column consists of an integer and the DMGEXP is either the letters K, M, or B correponding to “Thousands”, “Millions”, or “Billions”. Since we want to work with integers, we’ll have to combine the two columns together to get a total dollar value for each observation. For observations that have nothing in the DMGEXP field, we’ll replace those with a value of 1.
Let’s substitute the letters K,M,B for their numerical representation, and ensure these columns are read as numeric. We’ll then add a new column callled TOTALPROP and TOTALCROP which multiples the two columns. This will equal the total dollar value of damages for each event.
Dat$PROPDMGEXP <- gsub("K","1000",Dat$PROPDMGEXP)
Dat$PROPDMGEXP <- gsub("M","10000",Dat$PROPDMGEXP)
Dat$PROPDMGEXP <- gsub("B","1000000000",Dat$PROPDMGEXP)
Dat$PROPDMGEXP[Dat$PROPDMGEXP ==""] <- "1"
Dat$PROPDMGEXP <- as.numeric(Dat$PROPDMGEXP)
Dat$CROPDMGEXP <- gsub("K","1000",Dat$CROPDMGEXP)
Dat$CROPDMGEXP <- gsub("M","1000000",Dat$CROPDMGEXP)
Dat$CROPDMGEXP <- gsub("B","1000000000",Dat$CROPDMGEXP)
Dat$CROPDMGEXP[Dat$CROPDMGEXP ==""] <- "1"
Dat$CROPDMGEXP <- as.numeric(Dat$CROPDMGEXP)Dat$TOTALPROP <- (Dat$PROPDMGEXP*Dat$PROPDMG)
Dat$TOTALCROP <- (Dat$CROPDMGEXP*Dat$CROPDMG)Now that we have the total dollar amount of damage to both property and crops, let’s take a quick summary() of these new columns.
summary(Dat$TOTALPROP)## Min. 1st Qu. Median Mean 3rd Qu.
## 0 0 0 319200 500
## Max. NA's
## 115000000000 28summary(Dat$TOTALCROP)## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0 0 0 54410 0 5000000000
## NA's
## 29It appears that a high percentage of events carry no monetary damages, and very few events have high damages. To continue, we’ll just look at the 99th percentile in new data frames hiprop and hicrop.
hiprop <- subset(Dat,Dat$TOTALPROP >= (quantile(Dat$TOTALPROP,.99,na.rm=TRUE)))
hicrop <- subset(Dat,Dat$TOTALCROP >= (quantile(Dat$TOTALCROP,.99,na.rm=TRUE)))Finally, we’ll use ddply to create summary tables for property and crop damage totals, and specify we want to look at the total damages and mean damages per event type.
propsum <- ddply(hiprop,.(EVTYPE),summarise,total=sum(TOTALPROP),mean=mean(TOTALPROP))
propsum <- arrange(propsum,desc(total))
cropsum <- ddply(hicrop,.(EVTYPE),summarise,total=sum(TOTALCROP),mean=mean(TOTALPROP))
cropsum <- arrange(cropsum,desc(total))Results:
When examining events that pose a threat to population health, we should observe which types of disaster events have historically caused high fatalities and/or injuries. The graphs below depict fatalities and injuries for the most impactful types of disasters since 1950 (the earliest year of the dataset).
fatalplot1 <- ggplot(fatalsum[1:7,],aes(x=EVTYPE,y=mean,width=.7))+
geom_bar(stat="identity",colour="red",fill="red")+
xlab("Event Type")+
ylab("Mean Fatalities per Event")+
theme(axis.text.x = element_text(angle=-90, hjust=0.5, vjust=0.5,size=10,colour="black"))+
theme(axis.title.y = element_text(size=14,colour="black"))+
theme(axis.title.x = element_text(size=14,colour="black"))+
scale_x_discrete(limits=c("TORNADO","EXCESSIVE HEAT","FLASH FLOOD","HEAT","FLOOD","WILDFIRE","TSUNAMI"))
fatalplot2 <- ggplot(fatalsum[1:7,],aes(x=EVTYPE,y=total,width=.7))+
geom_bar(stat="identity",colour="red",fill="red")+
xlab("Event Type")+
ylab("Total Fatalities per Event")+
theme(axis.text.x = element_text(angle=-90, hjust=0.5, vjust=0.5,size=10,colour="black"))+
theme(axis.title.y = element_text(size=14,colour="black"))+
theme(axis.title.x = element_text(size=14,colour="black"))+
scale_x_discrete(limits=c("TORNADO","EXCESSIVE HEAT","FLASH FLOOD","HEAT","FLOOD","WILDFIRE","TSUNAMI"))injuryplot1 <- ggplot(injurysum[1:7,],aes(x=EVTYPE,y=mean,width=.7))+
geom_bar(stat="identity")+
xlab("Event Type")+
ylab("Mean Injuries per Event")+
theme(axis.text.x = element_text(angle=-90, hjust=0.5, vjust=0.5,size=10,colour="black"))+
theme(axis.title.y = element_text(size=14,colour="black"))+
theme(axis.title.x = element_text(size=14,colour="black"))+
scale_x_discrete(limits=c("TORNADO","EXCESSIVE HEAT","FLOOD","ICE STORM","HEAT","HURRICANE/TYPHOON","BLIZZARD"))
injuryplot2 <- ggplot(injurysum[1:7,],aes(x=EVTYPE,y=total,width=.7))+
geom_bar(stat="identity")+
xlab("Event Type")+
ylab("Total Injuries per Event")+
theme(axis.text.x = element_text(angle=-90, hjust=0.5, vjust=0.5,size=10,colour="black"))+
theme(axis.title.y = element_text(size=14,colour="black"))+
theme(axis.title.x = element_text(size=14,colour="black"))+
scale_x_discrete(limits=c("TORNADO","EXCESSIVE HEAT","FLOOD","ICE STORM","HEAT","HURRICANE/TYPHOON","BLIZZARD"))grid.arrange(fatalplot2,fatalplot1,ncol=2,main="Total and Mean Fatalites for Highest Fatality-Causing Events (Fig. 1)")
grid.arrange(injuryplot2,injuryplot1,ncol=2,main="Total and Mean Injuries for Highest Injury-Causing Events (Fig. 2)")
The graphs report several key insights about which events pose the largest threats:
- Tornados are by far the most common disaster associated with death or injury.
- Hurricanes and floods, while not being associated with significant fatalities, lead to a high number of injuries on average (400 and 498 injuries, respectively.)
- Heatwaves, flooding, and tsunamis, all which occur considerably less frequently than tornados, are responsible for a high number of deaths when they do occur.
Now we’ll examine which events had the most impactful economic consequences. Let’s combine the propsum and cropsum data frames to create graphs based on one data frame. We’ll add factor identifiers for “Property” and “Crop” so we can distinguish the ggplot facets.
propsum$type <- as.factor("Property")
cropsum$type <- as.factor("Crop")
alldamage <- rbind(head(propsum,30),head(cropsum,30))dmgplot1 <- ggplot(alldamage[1:60,],aes(x=EVTYPE,y=total/1000000000,width=.7))+
geom_bar(stat="identity")+
xlab("Event Type")+
ylab("Total Damage (in Billions of Dollars)")+
theme(axis.text.x = element_text(angle=-90, hjust=0.5, vjust=0.5,size=10,colour="black"))+
theme(axis.title.y = element_text(size=10,colour="black"))+
theme(axis.title.x = element_text(size=10,colour="black"))+
facet_grid(type ~.)+
theme(strip.background = element_rect(fill="orange"))+
scale_x_discrete(limits=c("DROUGHT","FLOOD","HURRICANE/TYPHOON","STORM SURGE","TORNADO","HURRICANE","TROPICAL STORM","WINTERSTORM","RIVERFLOOD","RIVER STORM"))dmgplot2 <- ggplot(alldamage[1:60,],aes(x=EVTYPE,y=mean/1000000000,width=.7))+
geom_bar(stat="identity")+
xlab("Event Type")+
ylab("Mean Damage (in Billions of Dollars)")+
theme(axis.text.x = element_text(angle=-90,size=7.5,colour="black"))+
theme(axis.title.y = element_text(size=10,colour="black"))+
theme(axis.title.x = element_text(size=10,colour="black"))+
facet_grid(type ~.)+
theme(strip.background = element_rect(fill="green"))+
scale_x_discrete(limits=c("HEAVY RAIN/SEVERE WEATHER","HURRICANE/TYPHOON","STORM SURGE","TORNADOES, TSTM WIND, HAIL","SEVERE THUNDERSTORM","HURRICANE OPAL","STORM SURGE/TIDE","RIVER FLOOD","HURRICANE","FLOOD"))Now we can arrange the two graphs together for the final figure 3:
grid.arrange(dmgplot1,dmgplot2,ncol=2,main="Total and Mean Economic Damages for Costliest Events (Fig. 3)")
A few conclusions about the nature of economic damages:
- Property damages largely outweigh agricultural damages in the United States. Total crop damages amount to less than 25 billion dollars since 1951, whereas property damages exceed 250 billion.
- Flooding and Storms are two categories which accounted for the most property damages since 1951. (Note that the dataset employs several descriptors for hurricanes and flooding.)
- Drought is the leading cause of agricultural damage, accounting for about half of all dollars in damages since 1951, but the mean drought occurance only costs $20,000, which reasons that drought events occur often.