I downloaded the data, and removed the columns I didn’t need. Then removed years before 1993 as there was little event data for those years. I converted the damange values from exponents into real values. Then I orderd the data and plotted the most impactful event types.
Load the required libraries
library(R.utils)
library(data.table)
library(R.cache)
library(dplyr)
library(lubridate)
library(ggplot2)
Downloading and Loading the Data, and only including the fields that will be used in analysis
download.file("https://d396qusza40orc.cloudfront.net/repdata%2Fdata%2FStormData.csv.bz2","noaa.csv.bz2", method="curl")
bunzip2("noaa.csv.bz2",remove=FALSE, skip=TRUE)
## [1] "noaa.csv"
## attr(,"temporary")
## [1] FALSE
noaa<-fread("noaa.csv",sep=",")[,c("BGN_DATE","EVTYPE", "FATALITIES", "INJURIES","PROPDMG","PROPDMGEXP","CROPDMG","CROPDMGEXP")]
Then I tidy up the data: change damange data into real numbers, change BGN to lubradate, create a total dmg column.
noaa<- noaa %>% mutate(date=mdy_hms(BGN_DATE))
noaa<- noaa %>% mutate(year=year(date))
noaa<-noaa %>% mutate(cropdmgv2 = case_when(CROPDMGEXP %in% c("B","b") ~ CROPDMG*100000000,
CROPDMGEXP %in% c("M","m") ~ CROPDMG*1000000,
CROPDMGEXP %in% c("K","k") ~ CROPDMG*1000,
CROPDMGEXP %in% c("H","h") ~ CROPDMG*100,
CROPDMGEXP %in% c("+") ~ CROPDMG*1,
CROPDMGEXP %in% c("-","?") ~ CROPDMG*0,
CROPDMGEXP %in% c(""," ") ~ CROPDMG*0,
CROPDMGEXP %in% 0:8 ~ CROPDMG*10,
))
noaa<-noaa %>% mutate(propdmgv2 = case_when(PROPDMGEXP %in% c("B","b") ~ PROPDMG*100000000,
PROPDMGEXP %in% c("M","m") ~ PROPDMG*1000000,
PROPDMGEXP %in% c("K","k") ~ PROPDMG*1000,
PROPDMGEXP %in% c("H","h") ~ PROPDMG*100,
PROPDMGEXP %in% c("+") ~ PROPDMG*1,
PROPDMGEXP %in% c("-","?") ~ PROPDMG*0,
PROPDMGEXP %in% c(""," ") ~ PROPDMG*0,
PROPDMGEXP %in% 0:8 ~ PROPDMG*10,
))
noaa<-noaa %>% mutate(totaldmg = propdmgv2+cropdmgv2)
noaa<-noaa %>% mutate(anydmg = propdmgv2+cropdmgv2+FATALITIES+INJURIES)
Then I determine which years have valueable data. In this case I find that it wasn’t until 1993 that there were a signifiant number of EV Types.
subset(noaa,year>1990 & year < 1995) %>%
group_by(year) %>%
summarise(count = n_distinct(EVTYPE))
## # A tibble: 4 x 2
## year count
## <dbl> <int>
## 1 1991 3
## 2 1992 3
## 3 1993 160
## 4 1994 267
I removed all years prior to 1993, plus any records where there were no damange or population health impact
noaav2<<-subset(noaa, anydmg>0 & year>1992)
By ordering the data by total fatalities and plotting i was able to determine which events are the most impactful to the population. The types of events that cause the greatest harm to human health are Tornado’s, Heat, and Floods
popimpact<-noaav2 %>%
group_by(EVTYPE) %>%
summarise(Total_Fatality = sum(FATALITIES), Total_Injury = sum(INJURIES), Total=sum(FATALITIES+INJURIES))
popimpact<-as.data.table(popimpact)
popimpact<-popimpact[order(-Total_Fatality)]
popimpact<-popimpact[1:10,]
popimpact<-melt(popimpact, id.vars="EVTYPE", variable.name="category")
g<-ggplot(popimpact, aes(x=reorder(EVTYPE,-value), y=value))+geom_bar(position="dodge",stat="identity", aes(fill=category))
g<-g+ylab("Count")+xlab("Event")+ggtitle("Top Population Impact") + theme(axis.text.x = element_text(angle=45, hjust=1))
print(g)
The by ordering the data by total damange I was able to determine which events have done the most damange. The types of events that cause the greatest economic consequences are Flood, Tornado’s, and Hail
dmg<-noaav2 %>%
group_by(EVTYPE) %>%
summarise(Prop_Dmg = sum(propdmgv2), Crop_Dmg = sum(cropdmgv2), Total=sum(propdmgv2+cropdmgv2))
dmg<-as.data.table(dmg)
dmg<-dmg[order(-Total)]
dmg<-dmg[1:10,]
dmg<-melt(dmg, id.vars="EVTYPE", variable.name="category")
g<-ggplot(dmg, aes(x=reorder(EVTYPE,-value), y=value))+geom_bar(position="dodge",stat="identity", aes(fill=category))
g<-g+ylab("Damage")+xlab("Event")+ggtitle("Top Crop and Properity Damage") + theme(axis.text.x = element_text(angle=45, hjust=1))
print(g)
