Peer Graded Assignment: Course Project 2
Sanjeeb Bhattarai
8/4/2020
#Peer Graded Assignment: Course Project 2 ##Population Health and Economic Effects of Weather Events Summary An analysis of NOAA Storm Database for effects of event types on population health and economic outcomes was conducted. The exploration of the data looked at fatalities, injuries, and total damage as a result of each weather event in the database. The analysis looks at both total fatalities, injuries, and damage by each event type and also looks at the mean effect of a single event for each event type. Objectives The basic goal of this analysis is to explore the NOAA Storm Database and answer two questions concerning severe weather events: -Which severe weather events (EVTYPE) are most harmful with respect to population health? -Which severe weather events have the greatest economic consequences? Sep-by-step method
- Data upload and familiarization Database was loaded into R using the read.csv function. Furthermore, the uploaded database was familiarised using str() and summary() commands.
#Reading data
dmg <- read.table("D:/knitR/R/repdata_data_StormData.csv", sep=",", header = TRUE)
str(dmg)
summary(dmg)- Data Preparation Time The recorded serve weather events started from January 1950 and ended at March 2016, but the coverage of recorded events changed over time: * Jan 1950 - Jan 1955 - Tornado * Jan 1955 - Jan 1966 - Tornado, Thunderstorm, Wind, and Hail * Jan 1966 - Mar 2016 - All Event Types Therefore data from Jan 1966 or later were kept for this analysis.
# Convert the date from factor to to a date field.
dmg$BGN_DATE <- as.Date(dmg$BGN_DATE, format = "%m/%d/%Y")
# Data as of January 1, 1966 is only taken for this analysis.
dmg <- subset(dmg, BGN_DATE >= as.Date("1966-01-01"))- Columns Due to the large size of columns, only necessary columns are kept for this analysis. BGN_DATE - Event Begin Date EVTYPE - Event Type FATALITIES - Number of reported fatalities caused by the event. INJURIES - Number of reported injuries caused by the event. PROPDMG/PROPDMGEXP - The dollar (USD) amount of property damage caused by the event. CROPDMG/CROPDMGEXP - The dollar (USD) amount of crop damage caused by the event.
# Necessary columns for the analysis which are kept for this analysis.
neededCols <- c("BGN_DATE",
"EVTYPE",
"FATALITIES",
"INJURIES",
"PROPDMG",
"PROPDMGEXP",
"CROPDMG",
"CROPDMGEXP")
# Take only cols and rows which are needed for this analysis.
data <- subset(dmg,
select = neededCols,
(grepl(EVTYPE, pattern = "*Summary*", ignore.case = TRUE) == FALSE))Adjustment of Several Event Type Names According to Storm Data Documentation(page 6), names must be written out, because some are represented in more than one name, such as “TSTM WIND” and “THUNDERSTORM WIND”.
data[data$EVTYPE == "TSTM WIND", ]$EVTYPE = "THUNDERSTORM WIND"
data[data$EVTYPE == "THUNDERSTORM WINDS", ]$EVTYPE = "THUNDERSTORM WIND"
data[data$EVTYPE == "RIVER FLOOD", ]$EVTYPE = "FLOOD"
data[data$EVTYPE == "HURRICANE/TYPHOON", ]$EVTYPE = "HURRICANE-TYPHOON"
data[data$EVTYPE == "HURRICANE", ]$EVTYPE = "HURRICANE-TYPHOON"- Conversion of Damage Amounts The damage amounts are stored in the data as a pair of columns. One is the “DMG” amount (PROPDMG/CROPDMG). The other is “DMGEXP”" which is the dimension to apply a magnitude to the amount (PROPDMGEXP/CROPDMGEXP). e.g. DMG x 10 ^ DMGEXP The distinct exponent symbols are identified so that they may be quantified individually.
# Create a dictanry for lookup exponents.
exponentLookup <- data.frame(DMGEXP = c("",
"-",
"?",
"+",
"0",
"1",
"2",
"3",
"4",
"5",
"6",
"7",
"8",
"B",
"H",
"K",
"M"
),
EXP = c(1, #
1, # -
1, # ?
1, # +
1, # 0
10, # 1
100, # 2
1000, # 3
10000, # 4
100000, # 5
1000000, # 6
10000000, # 7
100000000, # 8
1000000000, # B
100, # H
1000, # K
1000000 # M
),
stringsAsFactors = FALSE
)
# If DMGEXP has small letters convert them into capital letters
data$PROPDMGEXP <- toupper(data$PROPDMGEXP)
data$CROPDMGEXP <- toupper(data$CROPDMGEXP)
# Merge the power of exponents for use with property damage amounts.
data <- merge(x = data, y = exponentLookup,
by.x = "PROPDMGEXP", by.y = "DMGEXP",
all.x = TRUE)
names(data)[names(data)== "EXP"] <- "PROPEXP"
data <- merge(x = data, y = exponentLookup,
by.x = "CROPDMGEXP", by.y = "DMGEXP",
all.x = TRUE)
names(data)[names(data)== "EXP"] <- "CROPEXP"
# Apply the power of to the damage amounts
data$PropDamageAmount <- data$PROPDMG * data$PROPEX
data$CropDamageAmount <- data$CROPDMG * data$CROPEXP- Clean Texts of Event Type (EVTYPE)
#A general function "clean" of modifications to apply to all event type values.
clean <- function(x) {
x <- toupper(x) # use consistent casing
x <- sub("\\\\", " ", x) # eliminate \'s
x <- sub("-", " ", x) # eliminate -'s
x <- sub("/", " ", x) # eliminate /'s
x <- gsub("\\s+", " ", x) # compress multiple spaces
x <- trimws(x) # trim leading and trailing whitespace
}
data$EVTYPE <- apply(data.frame(data$EVTYPE), 2, clean)- Analysis The top 5 weather events, which have an impact to fatalities and injuries, will be ranked.
# Top 5 Severe Weather Events for Fatality
fatalityTop5 <- aggregate(list(Freq=data$FATALITIES),
list(EventType=data$EVTYPE),
sum)
fatalityTop5 <- head(fatalityTop5[order(-fatalityTop5$Freq),], n=5)
# Top 5 Severe Weather Events for Injury
injuryTop5 <- aggregate(list(Freq=data$INJURIES),
list(EventType=data$EVTYPE),
sum)
injuryTop5 <- head(injuryTop5[order(-injuryTop5$Freq),], n=5)library(ggplot2)
ggplot(data=fatalityTop5, aes(x=reorder(EventType, Freq), y=Freq)) +
geom_bar(stat="identity") +
ggtitle(label = "Top 5 Severe Weather Events with the Highest Total Fatalities") +
labs(x = "Event Type", y = "Count") +
coord_flip()
ggplot(data=injuryTop5, aes(x=reorder(EventType, Freq), y=Freq)) +
geom_bar(stat="identity") +
ggtitle(label = "Top 5 Severe Weather Events with the Highest Total Injuries") +
labs(x = "Event Type", y = "Count") +
coord_flip()
Tornados has at fatalities and injuries the highest number.
- Economic Consequences To analyze the top contributing weather events by total cost of damage, the total of both damage amounts (property and crop) will be calculated and ranked.
library(reshape)
# Calculte the costs of damage for property and for crop.
PropertyDamageTop5 <- aggregate(
list(PropDamageAmount=data$PropDamageAmount,
CropDamageAmount=data$CropDamageAmount),
list(EVTYPE=data$EVTYPE),
sum)
# Calculate the total costs
PropertyDamageTop5$TotalDamageAmount <- PropertyDamageTop5$PropDamageAmount + PropertyDamageTop5$CropDamageAmount
# Top 5 Contributing Severe Weather Events for Total Economic Damage
PropertyDamageTop5 <- PropertyDamageTop5[order(-PropertyDamageTop5$TotalDamageAmount),]
PropertyDamageTop5 <- head(PropertyDamageTop5, n=5)
# Reorganize PropertyDamageTop5 to 3 columns EVTYPE, variable, value
PropertyDamageTop5 <- melt(PropertyDamageTop5[,-4])## Using EVTYPE as id variablesnames(PropertyDamageTop5)[names(PropertyDamageTop5) == "variable"] <- "TypOfDamage"
names(PropertyDamageTop5)[names(PropertyDamageTop5) == "value"] <- "Amount"
# Rename PropDamageAmount and CropDamageAmount
levels(PropertyDamageTop5$TypOfDamage)[levels(PropertyDamageTop5$TypOfDamage) == "PropDamageAmount"] <- "Property Damage Amount"
levels(PropertyDamageTop5$TypOfDamage)[levels(PropertyDamageTop5$TypOfDamage) == "CropDamageAmount"] <- "Crop Damage Amount"library('scales')
ggplot(data=PropertyDamageTop5,
aes(x=reorder(EVTYPE, Amount), y=Amount, fill=TypOfDamage)) +
scale_fill_grey(start = 0, end = 0.9) +
theme(legend.position="bottom") +
guides(fill=guide_legend(title=NULL)) +
geom_bar(stat="identity", colour="black") +
ggtitle(label = "Top 5 Severe Weather Events with the Highest Economic Damage") +
labs(x = "Event Type", y = "Amount of Damage in USD") +
scale_y_continuous(labels = dollar) +
coord_flip()
Floods are the highest contributor economic loss. It is doubled of hurricane typhoon ,which is 2nd. Tornados, which is the highest contributor for fatalities and injuries, is just 3rd ranking event to economic damage. For these severe weather event types, costs for property damage is much higher than for crops damage.