Weather Impact on Health and Economy

Synoposis

The report is to download and analyse the data from National Weather Service according to two aspects. One is the health impact in terms of fatalities and injuries. The other is the economy impact in terms of property damage and crop damage.

Data Processing

• Downloaded data from this url : “https://d396qusza40orc.cloudfront.net/repdata%2Fdata%2FStormData.csv.bz2”

require(R.utils)
require(dplyr)
require(ggplot2)
destzipFileName = "stormData.bz2"
destCSV = "data.csv"
# Data Processing
if (!file.exists(destCSV))
{
    url = "https://d396qusza40orc.cloudfront.net/repdata%2Fdata%2FStormData.csv.bz2"
    download.file(url, destzipFileName, method="libcurl")
    bunzip2(destzipFileName, destCSV)
}
ds <- read.csv(destCSV, header=TRUE, sep=",")

• Pre-process the data

Using the dplyr package to group by the data with regarding their EVTYPE and sum up all the fatalties and injuries for health damage, and property/crop damages for economic impact. For the health damage we will just summarise the fatalties and injuries as a total.

ds <- read.csv(destCSV)
health_ds <- ds %>%
    group_by(EVTYPE) %>%
    summarise_at(vars(INJURIES, FATALITIES), sum) %>%
    arrange(desc(FATALITIES + INJURIES)) %>%
    mutate(health_damage = (FATALITIES + INJURIES)) %>%
    head(10)

For the economy damage, we will need to multiply the prop damage with the unit described in the PROPDMGEXP ,which has the code of K(1000), M(1 million), B (1 billion). There are some other code like 0,1,2…8 or ? , since those have limited impact on the data, we will not multily them.

economy_ds <- ds %>%
    select(EVTYPE, PROPDMG, CROPDMG, PROPDMGEXP, CROPDMGEXP) %>%
    mutate(PDMG=PROPDMG * case_when(PROPDMGEXP == "K" ~ 1000,
                                    (PROPDMGEXP == "M"| PROPDMGEXP == "m") ~ 1000000,
                                    PROPDMGEXP == "B" ~ 1000000000,
                                    TRUE ~ 1)) %>%
    mutate(CDMG=CROPDMG * case_when(CROPDMGEXP == "K" ~ 1000,
                                    (CROPDMGEXP == "M" | CROPDMGEXP == "m") ~ 1000000,
                                    CROPDMGEXP == "B" ~ 1000000000,
                                    TRUE ~ 1)) %>%
    group_by(EVTYPE) %>%
    summarise_at(vars(PDMG, CDMG), sum) %>%
    arrange(desc(PDMG+ CDMG))  %>%
    mutate(ECONOMY_DAMAGE= (PDMG+ CDMG) / 100000) %>%
    head(10)

Results

• Health damage: Only plot the top 10 weather disaster that caused the most damages. From plot, we can see the Tornado has been the greatest nature disaster in terms of health damage. The second most impact disaster is excessive heat.

require(ggplot2)
ggplot(aes(y=health_damage, x=reorder(EVTYPE,health_damage)), data=health_ds) + 
    geom_bar(stat="identity", aes(alpha=EVTYPE, col=EVTYPE, fill=EVTYPE)) + 
    coord_flip() + 
    ggtitle("Health Damange of Weather Disaster") + 
    labs(y="Sum of injuries and fatalties", x="Disaster Type")

• Economy damage: It turns out the flood damage most on economy, But the second most impact disaster is hurrican/typhoon, tornado the 3rd.

ggplot(aes(y=ECONOMY_DAMAGE, x=reorder(EVTYPE, ECONOMY_DAMAGE)), data=economy_ds) + 
    geom_bar(stat="identity", aes(alpha=EVTYPE,col=EVTYPE,fill=EVTYPE)) + 
    coord_flip() + 
    ggtitle("Economic Damange of Weather Disaster") + 
    labs(y="Sum of crop and properties damages (millions)", x="Disaster Type")