Analyze storm US data to calculate damages : humans and economics

Synopsis

The goal of the analyze is to compare events due to storm US like Tornado, Flood, Heat etc over the period : 1950 to 2011. First I analyze human damages (death and injuries) and I publish the top 10 events the more dangerous. Then I analyze the economics damages by weather disasters and again make a top 5 classification of the most important economic loss.

Data Processing

1. Download the file from : Storm Data
   
2. Unzip the file
   
3. Store the csv file
4. Change the RStudio working directory by the path where I store my file
   
5. Call my library package (for whom who don’t have already install package, use the function install.package(‘name’))

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

6. Load the data in a variable called “mdata” (Warning : data count 902 297 observations, it could be take a while)

mdata <- read.csv("StormData.csv")

Data Sorting

Working on the dataset and create subsets in variables for our different analyzes.

• Assign new datasets

md <- mdata %>% select(EVTYPE, FATALITIES,INJURIES) #subset for human damages
dmg <-  mdata %>% select(EVTYPE, PROPDMG , PROPDMGEXP) %>% filter(PROPDMG > 0) %>% group_by(EVTYPE) #subset for economics damages

• Create two variables which concern human damages (death and injuries)

dead <- md %>% filter(FATALITIES > 0) %>% group_by(EVTYPE) %>% summarise(FATALITIES = sum(FATALITIES)) %>% arrange(desc(FATALITIES))
hurt <- md %>% filter(INJURIES > 0) %>% group_by(EVTYPE) %>% summarise(INJURIES = sum(INJURIES)) %>% arrange(desc(INJURIES))

• Create one variable called “total_eco_dmg” which concern economic damages (death and injuries)

# Rework alphabetical characters used to signify magnitude include “K” for thousands, “M” for millions, and “B” for billions
dmg$PROPDMGEXP <- as.character(dmg$PROPDMGEXP)
dmg$PROPDMGEXP <- toupper(dmg$PROPDMGEXP)
dmg <- dmg %>% filter(PROPDMGEXP %in% c("B","K","M"))

# Create a new column and convert alpha character into numerical.
eco_dmg <- dmg %>% mutate(PROPDMGNUM = 10)
eco_dmg$PROPDMGNUM <- replace(eco_dmg$PROPDMGNUM,eco_dmg$PROPDMGEXP == "K",1000)
eco_dmg$PROPDMGNUM <- replace(eco_dmg$PROPDMGNUM,eco_dmg$PROPDMGEXP == "M",1000000)
eco_dmg$PROPDMGNUM <- replace(eco_dmg$PROPDMGNUM,eco_dmg$PROPDMGEXP == "B",1000000000)

# Calculate the total damage (need to make a product) and create a new variable which is clean for plot/screening
eco_dmg <- eco_dmg %>% mutate(Total_damage = PROPDMG*PROPDMGNUM)
total_eco_dmg <- eco_dmg %>% select(EVTYPE,Total_damage) %>% group_by(EVTYPE) %>% summarise(Total_damage = sum(Total_damage)/1e+09) %>% arrange(desc(Total_damage))
total_eco_dmg$Total_damage <- round(total_eco_dmg$Total_damage,2)

Results

Print the top 5 of our different analyzes :

1. Death by natural disasters

death <- head(dead,5)
barplot(death$FATALITIES,names.arg = death$EVTYPE, xlab='Weather event',ylab='Deaths',main='Top 5 of weather events causing death', col = c("#3399FF"))

2. Injury by natural disasters

hurt <- head(hurt,5)
barplot(hurt$INJURIES,names.arg = hurt$EVTYPE, xlab='Weather event',ylab='Injuries',main='Top 5 of weather events causing injuries', col = c("#3399FF") )

3. Economic losses by natural disasters (in Billion of dollars)

total_eco_dmg <- head(total_eco_dmg,5)
barplot(total_eco_dmg$Total_damage,names.arg = total_eco_dmg$EVTYPE, xlab='Weather event',ylab='in Billion of dollars',main='Top 5 of weather events causing economic losses', col = c("#3399FF") )