Most Destructive Weather Events in the United States 2006 Onwards

Synopsis

The data was read from the given .bz2 file and subset to get the data for the years 2006 onwards. The event type variable was cleaned as per the documentation to 48 event types. The effect on population health was gauged by the injuries and fatalities. The effect on economy, the crop and property damage were examined. In conclusion, tornadoes, excessive heat and thunderstorm wind were most harmful to population health, while flood, tornado and hail had the greatest economic consequences.

Loading Relevant Libraries

library(lubridate)

## 
## Attaching package: 'lubridate'

## The following objects are masked from 'package:base':
## 
##     date, intersect, setdiff, union

library(dplyr)

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:lubridate':
## 
##     intersect, setdiff, union

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

Data Processing

Reading Data

The data, along with its documentation, was downloaded from the provided URL and saved appropriately. The time of access is also recorded.

if(!file.exists("stormData.csv.bz2")) {
        url <- "https://d396qusza40orc.cloudfront.net/repdata%2Fdata%2FStormData.csv.bz2"
        download.file(url, "stormData.csv.bz2")
}

stormData <- read.csv("stormData.csv.bz2")
stormData <- as_tibble(stormData)

if(!file.exists("stormDataDocumentation.pdf")) {
        url2 <- "https://d396qusza40orc.cloudfront.net/repdata%2Fpeer2_doc%2Fpd01016005curr.pdf"
        download.file(url2, "stormDataDocumentation.pdf")
}

date.accessed <- Sys.time()
print(date.accessed)

## [1] "2020-05-15 15:37:48 IST"

Subsetting the Data

The data is very large and untidy. For convenience, for this project, a subset of the data has been considered. The data from 2006 onwards have been considered for further analysis.

The lubridate package was used to work with the dates. The variable names were also converted to the lowercase.

names(stormData) <- tolower(names(stormData))

# subsetting data 2006 onwards
isSelected <- year(mdy_hms(stormData$bgn_date)) > 2005

data <- stormData[isSelected, ]

Cleaning the Data

Fixing event types

The data initially had 54 unique entries. This wsa brought down to the 48 event types mentioned in the documentation. The strategy is to find a matching string in the evtype elements, and to rename that element properly.

evtype <- tolower(as.character(data$evtype))

# COASTAL FLOOD
isCoast <- grepl("^c", evtype)
isFlood <- grepl("flood", evtype)
evtype[isCoast & isFlood] <-  "coastal flood"
# JUST FLOOD
isLake <- grepl("lake", evtype)
isFlash <- grepl("flash", evtype)
evtype[isFlood & !(isCoast | isFlash | isLake)] <- "flood"


# COLD/WIND CHILL
isCold <- grepl("cold", evtype)
isExtreme <- grepl("extreme", evtype)
evtype[isCold & !isExtreme] <- "cold/wind chill"
# EXTREME COLD/WIND CHILL
evtype[isCold & isExtreme] <- "extreme cold/wind chill"

# DROUGHT
isDrought <- grepl("drought", evtype)
evtype[isDrought] <- "drought"

# DUST DEVIL
isDust <- grepl("dust", evtype)
isStorm <- grepl("storm", evtype)
evtype[isDust & !isStorm] <- "dust devil"

# FREEZING FOG
isFreeze <- grepl("freez", evtype)
isFog <- grepl("fog", evtype)
evtype[isFreeze & isFog] <- "freezing fog"
# DENSE FOG
evtype[isFog & !isFreeze] <- "dense fog"

# FROST/FREEZE
isFrost <- grepl("frost", evtype)
evtype[(isFreeze & !isFog) | isFrost] <- "frost/freeze"

# FUNNEL CLOUD
isFunnel <- grepl("funnel", evtype)
evtype[isFunnel] <- "funnel cloud"

# HAIL
isMarine <- grepl("marine", evtype)
isHail <- grepl("hail", evtype)
evtype[isHail & !isMarine] <- "hail"

# HEAT
isExcessive <- grepl("excess", evtype)
isHeat <- grepl("heat", evtype)
evtype[isHeat & !isExcessive] <-"heat"

# HIGH SURF
isSurf <- grepl("surf", evtype)
evtype[isSurf] <- "high surf"

# TSTM --> THUNDERSTORM
evtype <- sub("tstm", "thunderstorm", evtype)

# STORM TIDE
isTide <- grepl("tide|surge", evtype)
evtype[isStorm & isTide] <- "storm tide"

# WINTER WEATHER
isWeather <- grepl("weath", evtype)
evtype[isWeather] <- "winter weather"

# VOLCANIC ASH
isVolcano <- grepl("volc", evtype)
evtype[isVolcano] <- "volcanic ash"

# reassigning fixed evtype
data$evtype <- as.factor(evtype)

Fixing the exponents

The exponents are coded as K for thousands, M for millions and so on. They were converted to numeric format, i.e. K = 3, M = 6, and so on, as per the resources found on the discussion forum. Other than K, M and B, the other exponents were revalued as 0.

This was done with the help of a function ‘fixExp’ and a big ‘for’ loop.

fixExp <- function(x){
        isK <- grepl("K", x)
        isM <- grepl("M", x)
        isB <- grepl("B", x)
        if(isK){
                3
        } else if(isM){
                6
        } else if(isB){
                9
        } else {0}
}

reqData2 <- data %>%
                select(evtype, propdmg, propdmgexp, cropdmg, cropdmgexp)

reqData2$propdmgexp <- as.character(reqData2$propdmgexp)
reqData2$cropdmgexp <- as.character(reqData2$cropdmgexp)
for(i in 1:length(reqData2$propdmgexp)){
        reqData2$propdmgexp[i] <- fixExp(reqData2$propdmgexp[i])
        reqData2$cropdmgexp[i] <- fixExp(reqData2$cropdmgexp[i])
}

reqData2$propdmgexp <- as.numeric(reqData2$propdmgexp)
reqData2$cropdmgexp <- as.numeric(reqData2$cropdmgexp)

Results

Across the United States, which types of events are most harmful with respect to population health?

The injuries and fatalities were summed for each event type and the data set was ordered in their descending order. The dplyr package was used here.

reqData <- data %>%
                select(evtype, injuries, fatalities) %>%
                mutate(total = injuries + fatalities) %>%
                select(evtype, total) %>%
                group_by(evtype) %>%
                summarise(sum=sum(total))

reqData <- reqData[order(reqData$sum, decreasing=TRUE), ]

# barplot visualisation
par(mar=c(12, 4, 2, 1))
barplot(reqData$sum, names.arg=reqData$evtype, las=2, cex.names=0.7,
        ylab="Injuries and Fatalities",
        main="Effect of Weather Events on Population Health")

print(head(reqData))

## # A tibble: 6 x 2
##   evtype              sum
##   <fct>             <dbl>
## 1 tornado           11530
## 2 excessive heat     2197
## 3 thunderstorm wind  1775
## 4 heat               1451
## 5 lightning          1372
## 6 flash flood         686

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

Similarly, the crop and property damange were summed for each event type and the data set was reordered.

reqData2 <- reqData2 %>%
                mutate(crop=cropdmg*10^cropdmgexp, prop=propdmg*10^propdmgexp, total=crop + prop) %>%
                select(evtype, total) %>%
                group_by(evtype) %>%
                summarise(sum=sum(total))

reqData2 <- reqData2[order(reqData2$sum, decreasing=TRUE), ]

print(head(reqData2))

## # A tibble: 6 x 2
##   evtype                     sum
##   <fct>                    <dbl>
## 1 flood             133466686240
## 2 tornado            15491243740
## 3 hail                8401297600
## 4 flash flood         7914222260
## 5 storm tide          4642198000
## 6 thunderstorm wind   4210441770