Reproducible Research Peer Assessment 2: Analysis of NOAA Storm Database

Synopsis

This assignment explores and analyze the US NOAA Storm Database, which tracks major storms and weather events, to address the following two questions:

Across the United States, which types of events (as indicated in the EVTYPE variable) are most harmful with respect to population health?
Across the United States, which types of events have the greatest economic consequences?

Data Processing

This section contains reading in data, extracting only the data relevant to the questions, and cleaning the EVTYPE variable of the data.

Read in data.

storm=read.csv("repdata-data-StormData.csv.bz2",stringsAsFactors=FALSE)

Extract only the data that were collected after 1996, when 48 event types were appplied instead of 4 event types (more detail on event types).

storm$BGN_DATE=gsub(" 0:00:00","",storm$BGN_DATE)
storm$BGN_DATE=as.Date(storm$BGN_DATE,"%m/%d/%Y")
storm_96=subset(storm,BGN_DATE>="1996-01-01")

Extract only the variables relavant to the questions.

storm_relevant=with(storm_96,data.frame(BGN_DATE,EVTYPE,FATALITIES,INJURIES,PROPDMG,PROPDMGEXP,CROPDMG,PROPDMGEXP,stringsAsFactors=FALSE))

Extract only the events that were either harmful to population health (FATALITES or INJURIES>0) or having economic consequences (PROPDMG or CROPDMG>0).

storm_harmful=subset(storm_relevant,FATALITIES>0|INJURIES>0|PROPDMG>0|CROPDMG>0)

Append a new variable PROP to storm_harmful, which calculate the amount of property damage according to PROPDMG by PROPDMGEXP. If there is no value for ROPDMGEXP, assume it to be 1.

storm_harmful$PROP=storm_harmful$PROPDMG
for(i in 1:length(storm_harmful$PROPDMG)){
if(storm_harmful$PROPDMGEXP[i]=="K"){storm_harmful$PROP[i]=storm_harmful$PROPDMG[i]*1000}
if(storm_harmful$PROPDMGEXP[i]=="M"){storm_harmful$PROP[i]=storm_harmful$PROPDMG[i]*1000000}
if(storm_harmful$PROPDMGEXP[i]=="B"){storm_harmful$PROP[i]=storm_harmful$PROPDMG[i]*1000000000}
}

Append a new variable CROP to storm_harmful, which calculate the amount of crop damage according to CROPDMG by PROPDMGEXP.1. If there is no value for PROPDMGEXP.1, assume it to be 1.

storm_harmful$CROP=storm_harmful$CROPDMG
for(i in 1:length(storm_harmful$CROPDMG)){
if(storm_harmful$PROPDMGEXP.1[i]=="K"){storm_harmful$CROP[i]=storm_harmful$CROPDMG[i]*1000}
if(storm_harmful$PROPDMGEXP.1[i]=="M"){storm_harmful$CROP[i]=storm_harmful$CROPDMG[i]*1000000}
if(storm_harmful$PROPDMGEXP.1[i]=="B"){storm_harmful$CROP[i]=storm_harmful$CROPDMG[i]*1000000000}
}

Clean EVTYPE. When checked with length()andunique(), 222 event types are found in EVTYPE variable, much more than 48 event types suggested by the Storm Data Documentation, suggesting that there might be typos and other errors resulting in redundancy.

length(unique(storm_harmful$EVTYPE))

## [1] 222

Compare the 222 event types in storm_harmful with the 2.1.1 Storm Data Event Table in the Storm Data Documentation. Reduce the number of event types step by step.Here I only put the codes for this cleaning process, for more detail on the logic behind this cleaning, please check out this page.

storm_harmful$EVTYPE=toupper(storm_harmful$EVTYPE)
library(stringr)
storm_harmful$EVTYPE=str_trim(storm_harmful$EVTYPE)
storm_harmful[grep("COASTAL",storm_harmful$EVTYPE),]$EVTYPE="COASTAL FLOOD"
storm_harmful[grep("^COLD|HYPOTHERMIA",storm_harmful$EVTYPE),]$EVTYPE="COLD/WIND CHILL"
storm_harmful[grep(" COLD|WINDCHILL",storm_harmful$EVTYPE),]$EVTYPE="EXTREME COLD/WIND CHILL"
storm_harmful[grep("SLIDE|SLUMP",storm_harmful$EVTYPE),]$EVTYPE="DEBRIS FLOW"
storm_harmful[grep("BLOWING DUST",storm_harmful$EVTYPE),]$EVTYPE="DUST STORM"
storm_harmful[grep("HYPERTHERMIA|WARM",storm_harmful$EVTYPE),]$EVTYPE="HEAT"
storm_harmful[grep("HEAT WAVE|RECORD HEAT",storm_harmful$EVTYPE),]$EVTYPE="EXCESSIVE HEAT"
storm_harmful[grep("FLASH|DAM BREAK|HIGH WATER",storm_harmful$EVTYPE),]$EVTYPE="FLASH FLOOD"
storm_harmful[grep("RIVER",storm_harmful$EVTYPE),]$EVTYPE="FLOOD"
storm_harmful[grep("CSTL|TIDAL FLOODING",storm_harmful$EVTYPE),]$EVTYPE="COASTAL FLOOD"
storm_harmful[grep("ICE JAM",storm_harmful$EVTYPE),]$EVTYPE="FLASH FLOOD"
storm_harmful[grep("FROST|FREEZE",storm_harmful$EVTYPE),]$EVTYPE="FROST/FREEZE"
storm_harmful[grep("SMALL HAIL",storm_harmful$EVTYPE),]$EVTYPE="HAIL"
storm_harmful[grep("GUSTY WIND/HAIL",storm_harmful$EVTYPE),]$EVTYPE="HAIL"
storm_harmful[grep("GUSTY WIND/HVY RAIN|GUSTY WIND/RAIN",storm_harmful$EVTYPE),]$EVTYPE="HEAVY RAIN"
storm_harmful[grep("GUSTY WIND",storm_harmful$EVTYPE),]$EVTYPE="STRONG WIND"
storm_harmful[grep("FREEZING RAIN|RAIN/SNOW|FREEZING DRIZZLE",storm_harmful$EVTYPE),]$EVTYPE="WINTER WEATHER"
storm_harmful[grep("RAIN",storm_harmful$EVTYPE),]$EVTYPE="HEAVY RAIN"
storm_harmful[grep("EXCESSIVE SNOW|LATE SEASON SNOW|HEAVY SNOW SHOWER",storm_harmful$EVTYPE),]$EVTYPE="HEAVY SNOW"
storm_harmful[agrep("LAKE-EFFECT",storm_harmful$EVTYPE),]$EVTYPE="LAKE-EFFECT SNOW"
storm_harmful[grep("BLOWING SNOW|LIGHT SNOW|SNOW SQUALL|FALLING SNOW/ICE|SNOW AND ICE",storm_harmful$EVTYPE),]$EVTYPE="WINTER WEATHER"
storm_harmful$EVTYPE[which(storm_harmful$EVTYPE=="SNOW")]="HEAVY SNOW"
storm_harmful[grep("SURF",storm_harmful$EVTYPE),]$EVTYPE="HIGH SURF"
storm_harmful[grep("^HIGH WIND",storm_harmful$EVTYPE),]$EVTYPE="HIGH WIND"
storm_harmful[grep("HURRICANE|TYPHOON",storm_harmful$EVTYPE),]$EVTYPE="HURRICANE(TYPHOON)"
storm_harmful[grep("RIP|GRADIENT",storm_harmful$EVTYPE),]$EVTYPE="RIP CURRENT"
storm_harmful[grep("STORM SURGE",storm_harmful$EVTYPE),]$EVTYPE="STORM SURGE/TIDE"
storm_harmful[grep("ASTRONOMICAL HIGH",storm_harmful$EVTYPE),]$EVTYPE="COASTAL FLOOD"
storm_harmful[grep("^STRONG WIND",storm_harmful$EVTYPE),]$EVTYPE="STRONG WIND"
storm_harmful[grep("^THUNDERSTORM|^TSTM|BURST",storm_harmful$EVTYPE),]$EVTYPE="THUNDERSTORM WIND"
storm_harmful[grep("MARINE TSTM",storm_harmful$EVTYPE),]$EVTYPE="MARINE THUNDERSTORM WIND"
storm_harmful[agrep("NON-TSTM WIND",storm_harmful$EVTYPE),]$EVTYPE="OTHER"
storm_harmful[grep("TORNADO|LANDSPOUT",storm_harmful$EVTYPE),]$EVTYPE="TORNADO"
storm_harmful[grep("FIRE",storm_harmful$EVTYPE),]$EVTYPE="WILDFIRE"
storm_harmful[grep("WINTER WEATHER|BLACK ICE|GLAZE|ROAD|MIXED PRECIP|WINTRY|FREEZING SPRAY",storm_harmful$EVTYPE),]$EVTYPE="WINTER WEATHER"
storm_harmful[grep("^WIND|WHIRLWIND|WIND DAMAGE",storm_harmful$EVTYPE),]$EVTYPE="OTHER"
storm_harmful[grep("URBAN",storm_harmful$EVTYPE),]$EVTYPE="HEAVY RAIN"
storm_harmful[grep("DROWNING|MARINE ACCIDENT|BEACH EROSION",storm_harmful$EVTYPE),]$EVTYPE="OTHER"
storm_harmful[grep("SWELLS|ROGUE|SEAS",storm_harmful$EVTYPE),]$EVTYPE="HIGH SURF"

Results

1. Across the United States, which types of events (as indicated in the `EVTYPE` variable) are most harmful with respect to population health?

Top 10 event types causing most fatalities.

library(plyr)
fatalities=arrange(aggregate(FATALITIES~EVTYPE,data=storm_harmful,FUN=sum),FATALITIES,decreasing=TRUE)
head(fatalities,10)

##                     EVTYPE FATALITIES
## 1           EXCESSIVE HEAT       1799
## 2                  TORNADO       1511
## 3              FLASH FLOOD        890
## 4                LIGHTNING        651
## 5              RIP CURRENT        542
## 6                    FLOOD        416
## 7        THUNDERSTORM WIND        382
## 8  EXTREME COLD/WIND CHILL        258
## 9                     HEAT        238
## 10               HIGH WIND        235

When considering only fatalities, excessive heat is the most harmful event type, followed by tornado and flash flood.

Top 10 event types causing most injuries.

injuries=arrange(aggregate(INJURIES~EVTYPE,data=storm_harmful,FUN=sum),INJURIES,decreasing=TRUE)
head(injuries,10)

##                EVTYPE INJURIES
## 1             TORNADO    20667
## 2               FLOOD     6759
## 3      EXCESSIVE HEAT     6461
## 4   THUNDERSTORM WIND     5154
## 5           LIGHTNING     4141
## 6         FLASH FLOOD     1674
## 7            WILDFIRE     1458
## 8  HURRICANE(TYPHOON)     1328
## 9        WINTER STORM     1292
## 10               HEAT     1241

When considering only injuries, the most harmful event type is tornado, followed by flood and excessive heat. Plot top 10 events causing fatalities and that causing injuries into one stacked bar plot.

colnames(fatalities)=c("EVTYPE","NUMBER")
colnames(injuries)=c("EVTYPE","NUMBER")
top10_harmful=rbind(mutate(head(fatalities,10),factor=rep("FATALITIES",10)),mutate(head(injuries,10),factor=rep("INJURIES",10)))
library(ggplot2)
ggplot(top10_harmful,aes(x=EVTYPE,y=NUMBER,fill=factor))+
  geom_bar(position="stack",stat="identity")+
  labs(title="Event Types Most Harmful to Public Health",fill="",x="Event Type",y="Number of Fatalities/Injuries")+
  theme(axis.text.x=element_text(angle=30,hjust=1))

From this plot, we can see that when considering both the numbers of fatalities and injuries, tornado is the most harmful event type, followed by excessive heat and flood.

2. Across the United States, which types of events have the greatest economic consequences?

Top 10 event types having greatest economic consequences.

storm_harmful=mutate(storm_harmful,DAMAGE_IN_BILLION_USD=(PROP+CROP)/1000000000)
damage=aggregate(DAMAGE_IN_BILLION_USD~EVTYPE,data=storm_harmful,FUN=sum)
top10_damage=head(arrange(damage,DAMAGE_IN_BILLION_USD,decreasing=TRUE),10)
top10_damage

##                EVTYPE DAMAGE_IN_BILLION_USD
## 1  HURRICANE(TYPHOON)               1617.42
## 2               FLOOD                227.49
## 3             TORNADO                 49.92
## 4         FLASH FLOOD                 48.67
## 5    STORM SURGE/TIDE                 47.84
## 6                HAIL                 27.56
## 7   THUNDERSTORM WIND                 21.00
## 8            WILDFIRE                 15.03
## 9           HIGH WIND                 12.42
## 10     TROPICAL STORM                  9.46

Plot top 10 event types having greatest economic consequences.

ggplot(top10_damage,aes(x=EVTYPE,y=DAMAGE_IN_BILLION_USD))+
  geom_bar(stat="identity",width=0.75,fill="#009E73")+
  labs(title="Event Types with Greatest Economic Consequences",x="Event Type",y="Damage (in Billion USD)")+
  theme(axis.text.x=element_text(angle=30,hjust=1))

plot of chunk unnamed-chunk-14

From the table and plot above, we can see that hurricane(typhoon) is the event type having greatest economic consequences, followed by flood and tornado. Also, hurricane(typhoon) outnumbers flood in economic consequences by about 7 to 1.