Reproducible Research Project 2
Tracy Wormwood
2024-06-10
Analysis of health and monetary impacts of weather events in USA from 1950-2011.
Synopsis
The data analyzed in this report was extracted from the U.S. National Oceanic and Atmospheric Administration’s (NOAA) storm database. The data collected details fatalities and injuries to humans, as well as crop and property damage for more than 900,000 weather events over a time period of 60 years in the United States. In order to determine which weather phenomena have been most harmful to human health, the total injuries and fatalities for each weather event type were calculated, and the event types with the highest sums are reported. To determine which weather phenomena were most costly, the total costs of crop and property damages were calculated and the event types with the highest sums are reported here.
Data Processing
Download and impact on health
Data can be downloaded from a bzip file. The csv is extracted upon download automatically.
fileUrl<-"https://d396qusza40orc.cloudfront.net/repdata%2Fdata%2FStormData.csv.bz2"
download.file(fileUrl, destfile="/Users/Tracy/Desktop/RR Project 2/data.csv")Next, the data is read into R. To investigate the weather events that are most harmful to human health (Question 1), the dataset will be cropped to only include the variables event type, fatalities and injuries.
data<-read.csv("./data.csv")
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, unionhealth<-select(data, EVTYPE, FATALITIES, INJURIES)
event_cat<-unique(health$EVTYPE)
table_event<-table(health$EVTYPE)
table_event<-as.data.frame(table_event)Next, the sum of fatalities and injuries for each event type are calculated using tapply:
sum_fatal<-tapply(health$FATALITIES, health$EVTYPE, sum)
sum_fatal<-as.data.frame(sum_fatal)
sum_fatal$evtype<-table_event$Var1
sum_injury<-tapply(health$INJURIES, health$EVTYPE, sum)
sum_injury<-as.data.frame(sum_injury)
sum_injury$evtype<-table_event$Var1After those sums are calculated, the top ten event causes for both injuries and fatalities are reported, and then intersect() can be used to find which event types are in common between the two datasets (injury and fatality).
sum_fatal_num<-filter(sum_fatal, sum_fatal>0)
sum_injury_num<-filter(sum_injury, sum_injury>0)
order_fatal<-sum_fatal[with(sum_fatal, order(-sum_fatal)),]
top_10_fatal<-order_fatal[1:10,]
order_inj<-sum_injury[with(sum_injury, order(-sum_injury)),]
top_10_injury<-order_inj[1:10,]
common<-intersect(top_10_fatal$evtype, top_10_injury$evtype)
common## [1] "TORNADO" "EXCESSIVE HEAT" "FLASH FLOOD" "HEAT"
## [5] "LIGHTNING" "TSTM WIND" "FLOOD"Processing for economic impact
In order to process data to answer question 2, only the required variables are kept in the dataset. These are event type, property damage, property damage exponent, crop damage, and crop damage exponent. The values listed for property damage exponent appear to be confusing so this will be investigated here.
damage<-select(data, EVTYPE, PROPDMG, PROPDMGEXP, CROPDMG, CROPDMGEXP)
table(damage$PROPDMGEXP)##
## - ? + 0 1 2 3 4 5 6
## 465934 1 8 5 216 25 13 4 4 28 4
## 7 8 B h H K m M
## 5 1 40 1 6 424665 7 11330The PROPDMGEXP value is supposed to be in K (1000), M (1000000), or B (1000000000),according to the Storm Data Prep file. However, as shown in the code chunk above, different values are also input in this column. Some, the numbers, could be exponents, however the majority of the values are in K, M, or B. Do those labeled with “?” have a property damage numeric value?
filter(damage, PROPDMGEXP=="?")## EVTYPE PROPDMG PROPDMGEXP CROPDMG CROPDMGEXP
## 1 THUNDERSTORM WINDS 0 ? 0
## 2 FLASH FLOOD 0 ? 0
## 3 FLASH FLOOD 0 ? 0
## 4 THUNDERSTORM WIND 0 ? 0
## 5 HAIL 0 ? 0
## 6 HAIL 0 ? 0
## 7 HAIL 0 ? 0
## 8 THUNDERSTORM WINDS 0 ? 0All of these rows have “0” value for property damage. These can be filtered out. Additionally, do those labeled with “+” for the property damage exponent have a value?
filter(damage, PROPDMGEXP=="+")## EVTYPE PROPDMG PROPDMGEXP CROPDMG CROPDMGEXP
## 1 BREAKUP FLOODING 20 + 0
## 2 HIGH WIND 20 + 0
## 3 FLOODING/HEAVY RAIN 2 + 0
## 4 HIGH WINDS 15 + 0
## 5 TORNADO 60 + 0Here, the five rows with a “+” do have a numeric value. I’m going to read the notes from a few of these cases, but first an index must be added to the damage data frame so the case can be easily looked up from the data data frame.
damage$index<-seq.int(nrow(damage))
plus<-filter(damage, PROPDMGEXP=="+")The case notes from index=188780:
data[188780,]## STATE__ BGN_DATE BGN_TIME TIME_ZONE COUNTY COUNTYNAME STATE
## 188780 2 5/1/1995 0:00:00 0000 AST 0 AKZ001 AK
## EVTYPE BGN_RANGE BGN_AZI BGN_LOCATI END_DATE
## 188780 BREAKUP FLOODING 0 Southwest Alaska 5/31/1995 0:00:00
## END_TIME COUNTY_END COUNTYENDN END_RANGE END_AZI END_LOCATI LENGTH WIDTH
## 188780 0 NA 0 0 0
## F MAG FATALITIES INJURIES PROPDMG PROPDMGEXP CROPDMG CROPDMGEXP WFO
## 188780 NA 0 0 0 20 + 0
## STATEOFFIC ZONENAMES LATITUDE LONGITUDE LATITUDE_E LONGITUDE_
## 188780 HOUSTON - HOUSTON 0 0 0 0
## REMARKS
## 188780 Typical of this time of year is the seasonal breakup process across mainland Alaska. Major rivers, which serve as transportation corridors around the state, undergo the change from ice covered roadways (even plowed by the state) to summer boat/barge traffic. May 6th and 7th brought flooding to Bethel, on the Kuskokwim River. Damage occurred to the seawall and caused Brown Slough to back up. Areas along the waterfront and low areas bordering the slough were flooded, worst of which occurred along Alder Street, East Avenue, Fourth through Seventh Avenues, Main Street, and Settler Drive. Several vehicles were nearly submerged and several people were rescued from homes that had flood water coming in. Additionally, there were several oil spills from local tanks used to heat homes. Other villages affected by breakup on the Kuskokwim River were (in order) Aniak (damaged roads and damaged dike), Akiak, Akiachak, Kwethluk (airport road damage, extensive boardwalk damage, and insulation damage to 13 homes), Oscarville (flooded sewage lagoon, flooded streets and boardwalk, and a few flooded structures), and then Napaskiak (flooded roads and the new runway was isolated, insulation damage to six homes). Upper and Lower Kalskag were also affected by "minor" breakup flooding. Minor flooding was reported along the Yukon River at Pilot Station, Marshall, Emmonak (six houses), and Alakanuk (roads and other low lying areas).
## REFNUM
## 188780 188780The property damage value is 20, and the exponent is “+”. From reading the notes it’s not clear how the number 20 was determined. To make this analysis simpler, the damage data frame will be filtered to only include cases which have a K, M, or B for the property damage exponent value. The same will also be applied for crop damage exponent values. Then, lowercase values will be converted to uppercase values.
list<-c("K", "k", "M", "m", "B")
damage_filter<-filter(damage, PROPDMGEXP %in% list | CROPDMGEXP %in% list)
damage_filter$PROPDMGEXP2<-toupper(damage_filter$PROPDMGEXP)
damage_filter$CROPDMGEXP2<-toupper(damage_filter$CROPDMGEXP)The uppercase values were created in a new column. The old column will be removed and the “new” column will be renamed. Additionally, there appear to be some cases where there is a “K” value, but the number for damage is 0. These need to be removed as well (damage_filter_both).
damage_filter<-select(damage_filter, EVTYPE, PROPDMG, PROPDMGEXP2, CROPDMG, CROPDMGEXP2, index)
damage_filter<-rename(damage_filter, PROPDMGEXP=PROPDMGEXP2, CROPDMGEXP=CROPDMGEXP2)
damage_filter_both<-filter(damage_filter, PROPDMG>0 | CROPDMG>0)Now the damage values should be calculated in numeric form. The property or crop damage value should be multiplied by the “K”, “M” or “B” value and should be stored in a new column. This can be performed by writing two functions, multiplier_prop and multiplier_crop:
multiplier_prop<-function(dataset) {
newdataset<-data.frame()
for (i in 1:nrow(dataset))
if (dataset$PROPDMGEXP[[i]]=="K") {
newdataset<-rbind(newdataset, value=dataset$PROPDMG[[i]]*1000)}
else if (dataset$PROPDMGEXP[[i]]=="M") {
newdataset<-rbind(newdataset, value=dataset$PROPDMG[[i]]*1000000)}
else if (dataset$PROPDMGEXP[[i]]=="B") {
newdataset<-rbind(newdataset, value=dataset$PROPDMG[[i]]*1000000000)}
else newdataset<-rbind(newdataset, value=0)
newdataset
}
multiplier_crop<-function(dataset) {
newdataset<-data.frame()
for (i in 1:nrow(dataset))
if (dataset$CROPDMGEXP[[i]]=="K") {
newdataset<-rbind(newdataset, value=dataset$CROPDMG[[i]]*1000)}
else if (dataset$PROPDMGEXP[[i]]=="M") {
newdataset<-rbind(newdataset, value=dataset$CROPDMG[[i]]*1000000)}
else if (dataset$PROPDMGEXP[[i]]=="B") {
newdataset<-rbind(newdataset, value=dataset$CROPDMG[[i]]*1000000000)}
else newdataset<-rbind(newdataset, value=0)
newdataset
}Then, the output of these two functions was added to the damage_filter_both dataset. These are numeric values for property and crop damage.
damage_filter_both$propertydamage<-multiplier_prop(damage_filter_both)
damage_filter_both$cropdamage<-multiplier_crop(damage_filter_both)After these values are calculated, tapply can be run to find the sum of property/crop damages by event type. The two resulting data frames are summed in the dataset (both) to determine the total economic impact.
prop_sum<-tapply(damage_filter_both$propertydamage, damage_filter_both$EVTYPE, sum)
prop_sum<-as.data.frame(prop_sum)
crop_sum<-tapply(damage_filter_both$cropdamage, damage_filter_both$EVTYPE, sum)
crop_sum<-as.data.frame(crop_sum)
both<-cbind(crop_sum, prop_sum)
both$eventtype<-rownames(both)
both$total<-(crop_sum+prop_sum)As with human health, it would be interesting to find the top 10 event types regarding economic impact for both crop and property damage. It would also be interesting to know the sum of crop and property damages for event types and see which events are most costly.
order_prop<-both[with(both, order(-prop_sum)),]
top_10_prop<-order_prop[1:10,]
order_crop<-both[with(both, order(-crop_sum)),]
top_10_crop<-order_crop[1:10,]
common_value<-intersect(top_10_crop$eventtype, top_10_prop$eventtype)
common_value## [1] "HURRICANE" "HURRICANE/TYPHOON" "FLOOD"
## [4] "HAIL" "FLASH FLOOD"Earlier the sum of crop and property damage (both$total) was calculated. Now these values can be ordered and the top ten can be reported. The sum calculated earlier is recorded as a dataframe so it must be transformed that to an object that can be ordered.
str(both)## 'data.frame': 426 obs. of 4 variables:
## $ crop_sum : num 0 0 0 0 0 0 0 0 0 0 ...
## $ prop_sum : num 200000 50000 8100000 8000 5000 ...
## $ eventtype: chr " HIGH SURF ADVISORY" " FLASH FLOOD" " TSTM WIND" " TSTM WIND (G45)" ...
## $ total :'data.frame': 426 obs. of 1 variable:
## ..$ crop_sum: num 200000 50000 8100000 8000 5000 ...total_value<-select(both, 3:4)
str(total_value)## 'data.frame': 426 obs. of 2 variables:
## $ eventtype: chr " HIGH SURF ADVISORY" " FLASH FLOOD" " TSTM WIND" " TSTM WIND (G45)" ...
## $ total :'data.frame': 426 obs. of 1 variable:
## ..$ crop_sum: num 200000 50000 8100000 8000 5000 ...total_value_2<-unlist(total_value$total)
str(total_value_2)## Named num [1:426] 200000 50000 8100000 8000 5000 ...
## - attr(*, "names")= chr [1:426] "crop_sum1" "crop_sum2" "crop_sum3" "crop_sum4" ...both$totalval<-cbind(total_value_2)
order_total<-both[with(both, order(-totalval)),]
top_10_total<-order_total[1:10,]
top_10_total## crop_sum prop_sum eventtype
## HURRICANE 802881910000 11868319010 HURRICANE
## HURRICANE/TYPHOON 430370342800 69305840000 HURRICANE/TYPHOON
## FLOOD 35149748450 144657709800 FLOOD
## TORNADO 194933110 56937160480 TORNADO
## STORM SURGE 5000 43323536000 STORM SURGE
## FLASH FLOOD 1058517100 16140811510 FLASH FLOOD
## HAIL 1127274450 15732266720 HAIL
## WILDFIRE 6593472800 4765114000 WILDFIRE
## RIVER FLOOD 5022459000 5118945500 RIVER FLOOD
## HURRICANE OPAL/HIGH WINDS 10000000000 100000000 HURRICANE OPAL/HIGH WINDS
## crop_sum total_value_2
## HURRICANE 814750229010 814750229010
## HURRICANE/TYPHOON 499676182800 499676182800
## FLOOD 179807458250 179807458250
## TORNADO 57132093590 57132093590
## STORM SURGE 43323541000 43323541000
## FLASH FLOOD 17199328610 17199328610
## HAIL 16859541170 16859541170
## WILDFIRE 11358586800 11358586800
## RIVER FLOOD 10141404500 10141404500
## HURRICANE OPAL/HIGH WINDS 10100000000 10100000000common_total<-intersect(top_10_total$eventtype, common_value)
common_total## [1] "HURRICANE" "HURRICANE/TYPHOON" "FLOOD"
## [4] "FLASH FLOOD" "HAIL"Results
Across the United States, which types of events (as indicated in the EVTYPE variable) are most harmful with respect to population health?
The weather events that are most damaging to human health, considering fatalities and injuries, are TORNADO, EXCESSIVE HEAT, FLASH FLOOD, HEAT, LIGHTNING, TSTM WIND, FLOOD.
The package ggplot2 is used to create a bar graph showing the event types with the most fatalities and injuries.
library(ggplot2)
f<-ggplot(top_10_fatal, aes(x=evtype, y=sum_fatal))+geom_col(fill="pink")+geom_text(aes(label=evtype), size=2, vjust=-0.3)+ylab("Total fatalities")+xlab("Event type")+ggtitle("Weather events with most fatalities 1950-2011 (USA)")+theme(axis.text.x=element_blank(), axis.ticks = element_blank(), legend.position = "none")
i<-ggplot(top_10_injury, aes(x=evtype, y=sum_injury))+geom_col(fill="lightblue")+geom_text(aes(label=evtype), size=2, vjust=-0.3)+ylab("Total injuries")+xlab("Event type")+ggtitle("Weather events with most injuries 1950-2011 (USA)")+theme(axis.text.x=element_blank(), axis.ticks = element_blank(), legend.position = "none")
library(patchwork)
f/i
png(filename = "Healthimpact.png", width=480, height=960)
f/i
dev.off()## quartz_off_screen
## 2The graph shows that tornadoes are by far the most harmful to human health.
Across the United States, which types of events have the greatest economic consequences?
There are five event types in common for the most costly damages. These five are HURRICANE, HURRICANE/TYPHOON, FLOOD, HAIL, FLASH FLOOD.
Considering the total of property and crop damage, the most costly weather event types are HURRICANE, HURRICANE/TYPHOON, FLOOD, FLASH FLOOD, HAIL.
Finally, a plot will be created to summarize the investigation of question 2.
library(data.table)##
## Attaching package: 'data.table'## The following objects are masked from 'package:dplyr':
##
## between, first, lasttop_10_t<-as.data.table(top_10_total)
top_10_damage<-melt(top_10_t, id.vars=3, measure.vars=c(1,2))
dam<-ggplot(top_10_damage, aes(x=eventtype, y=value, fill=variable))+geom_col()+theme(axis.text.x = element_text(size=7, angle=35))+labs(x="Weather event type", y="Total in USD", title="Economic Damages by Weather Event 1950-2011 (USA)")+scale_fill_discrete(name="Asset type", labels = c("Crop", "Property"))
dam
png(filename="Economicimpact.png", width=900, height=600)
dam
dev.off()## quartz_off_screen
## 2The crop damage incurred by hurricanes and typhoons is notably higher than the other weather event types. Floods have are costliest in terms of property damage.
Conclusion
In summary, from 1950-2011 in the United States tornadoes were most harmful to human health, and hurricanes had the greatest economic consequences.