Analysis of health impact and economic damages from weather events in USA

Synopsis

The purpose of data analysis is to determine the impact of NOAA weather events on the health of the population and economic losses. The analysis showed that the greatest impact on the health and economic losses generate tornadoes and flash flood and thunderstorm wind. The data used in the analysis come from the years 1950-2011. Analysis suggests that the severity of the impact on health and economic losses increases from the 90s of the twentieth century. However, this is likely to result from the fact that in previous years were not collected as accurately as current data on weather phenomena.

Data Processing

First we need load two R packages (dplyr and ggplot2) which I will use in my analisys.

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
library("ggplot2")

Downlading data Storm data from working directory

download.file("https://d396qusza40orc.cloudfront.net/repdata%2Fdata%2FStormData.csv.bz2", destfile="stormData.csv.bz2", method="curl")
## Warning: uruchomione polecenie 'curl  "https://d396qusza40orc.cloudfront.net/repdata%2Fdata%2FStormData.csv.bz2"  -o "stormData.csv.bz2"' otrzymało status 127
## Warning: download had nonzero exit status
data <- read.table("storm1.csv", header=T, sep = ",")

Creating a data frame from the selected variables, which will be used for analysis The new data frame will contain the variables: BGN_DATE, EVTYPE, FATALITIES, INJURIES, PROPDMG, CROPDMG

data.selected <- select(data, BGN_DATE, EVTYPE,FATALITIES, INJURIES, PROPDMG,CROPDMG)

BGN_DATE variable must be converted to Dates class. This is done by replacing BGN_DATE using as.character and then with the function strptime to POSIXlt. At the end I create a new variable Year from BGN_DATE variable

data.selected$BGN_DATE <- strptime(as.character(data.selected$BGN_DATE),"%m/%d/%Y") 
data.selected$YEAR <- format(data.selected$BGN_DATE,"%Y")  
data.selected$BGN_DATE <- as.Date(data.selected$BGN_DATE)

Let look at the number of weather events in different years

hist(as.numeric(data.selected$YEAR), xlab="Year", main="Number of weather events per year")

plot of chunk unnamed-chunk-5

Variable EVTYPE requires little cleaning. First of all, I delete the values in which the’re is a string “Summaries.”

remove.summary <- grep("Summary.", data.selected[, "EVTYPE"], ignore.case=TRUE)
data.selected <- data.selected[-remove.summary, ]

Results

Summarise of health and economic impact data. We summarise two health related variables: FATALITIES and INJURIES and economic variables: PROPDMG and CROPDMG.

data.selected<-mutate(data.selected, HEALTH.HARMFUL=FATALITIES+INJURIES)
data.selected<-mutate(data.selected, ECONOMIC.DAMAGES=PROPDMG+CROPDMG)

Group data by Weather events type and by year

data.selected.by.EVTYPE <- group_by(data.selected, EVTYPE)  
data.selected.by.YEAR <- group_by(data.selected,YEAR)

Summarise variables by weather events.

data.selected.sum <- summarise(data.selected.by.EVTYPE, count=n(), injuries=sum(INJURIES), 
                               fatalities=sum(FATALITIES),prop.damages=sum(PROPDMG),crop.damages=sum(CROPDMG),
                               health=sum(HEALTH.HARMFUL), economic.damages=sum(ECONOMIC.DAMAGES))
data.selected.sum.by.YEAR <- summarise(data.selected.by.YEAR, count=n(), injuries=sum(INJURIES), 
                                      fatalities=sum(FATALITIES),prop.damages=sum(PROPDMG),crop.damages=sum(CROPDMG),
                                      health=sum(HEALTH.HARMFUL), economic.damages=sum(ECONOMIC.DAMAGES))
  1. Across the United States, which types of events (as indicated in the EVTYPE variable) are most harmful with respect to population health?

Creating health impact ranking

data.selected.ranking.health <- arrange(data.selected.sum, desc(health))

First ten highest public health impact

data.selected.ranking.health[1:10,c("EVTYPE","health","injuries","fatalities")]
## Source: local data frame [10 x 4]
## 
##               EVTYPE health injuries fatalities
## 1            TORNADO  96979    91346       5633
## 2     EXCESSIVE HEAT   8428     6525       1903
## 3          TSTM WIND   7461     6957        504
## 4              FLOOD   7259     6789        470
## 5          LIGHTNING   6046     5230        816
## 6               HEAT   3037     2100        937
## 7        FLASH FLOOD   2755     1777        978
## 8          ICE STORM   2064     1975         89
## 9  THUNDERSTORM WIND   1621     1488        133
## 10      WINTER STORM   1527     1321        206

It’s clear, that most public health impact is from tornadoes.

Look how it was health impact year by year

plot(data.selected.sum.by.YEAR$YEAR, data.selected.sum.by.YEAR$health, xlab="Year", ylab="Health impact", main="Yearly health impact from weather events")

plot of chunk unnamed-chunk-12

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

Creating economic damages ranking

data.selected.ranking.economic <- arrange(data.selected.sum, desc(economic.damages))

First ten highest economic damages

data.selected.ranking.economic[1:10,c("EVTYPE","economic.damages","prop.damages","crop.damages")]
## Source: local data frame [10 x 4]
## 
##                EVTYPE economic.damages prop.damages crop.damages
## 1             TORNADO          3312277      3212258       100019
## 2         FLASH FLOOD          1599325      1420125       179200
## 3           TSTM WIND          1445168      1335966       109203
## 4                HAIL          1268290       688693       579596
## 5               FLOOD          1067976       899938       168038
## 6   THUNDERSTORM WIND           943636       876844        66791
## 7           LIGHTNING           606932       603352         3581
## 8  THUNDERSTORM WINDS           464978       446293        18685
## 9           HIGH WIND           342015       324732        17283
## 10       WINTER STORM           134700       132721         1979

Tornadoes has also a bigest ecomomic impact.

Look at economic damages year by year

plot(data.selected.sum.by.YEAR$YEAR, data.selected.sum.by.YEAR$economic.damages, xlab="Year", ylab="Economic damages", main="Yearly economic damages from weather events")

plot of chunk unnamed-chunk-15

Summary of results

  • It seems that tornadoes have the highest health impact.
  • Tornadoes have also the highest economic impact in terms of propertyand crop damage.
  • It seems that the number of incidents, weather increases in recent years. Most likely due to the fact that previously were not collected them so thoroughly.