Thetazero Pubs
iscyangdata storm storms impacts billions billion events event casualty injury weathers

Storm-related disaster and its social and economical impacts

Synopsis

Even though scientists are still arguing whether global warming exists and if the atmospheric temperature over last few decades was really increasing, it is apparent to us that frequency of severe weathers and record-breaking temperatures is getting higher each year. Those events often cause casualties, injuries and various types of damages, namely property and crop damages.


Therefore, we obtained storm data, collected and published by National Oceanic and Atmospheric Administration (NOAA) of United States Department of Commerce, from 1950 to November of 2011. We then analyzed the data in order to obtain solid evidences on the social and economical impacts caused by different types of storm-related events.

From the above data and through our analysis, we found that tornado caused most casualties and injuries while flood resulted in most property and crop damages.

Data Processing

First, we load the NOAA storm data into variable ns.

ns = read.csv("repdata-data-StormData.csv.bz2", stringsAsFactor = F)

We have to convert property damages into numeric by combining its exponent term and mantissa.

ns$propdmge = ns$PROPDMGEXP
ns$propdmge[tolower(ns$propdmge) == "b"] = "1000000000"
ns$propdmge[tolower(ns$propdmge) == "m"] = "1000000"
ns$propdmge[tolower(ns$propdmge) == "k"] = "1000"
ns$propdmge[tolower(ns$propdmge) == "h"] = "100"
ns$propdmge[tolower(ns$propdmge) == "+"] = "1"
ns$propdmge[tolower(ns$propdmge) == "?"] = "1"
ns$propdmge[tolower(ns$propdmge) == "-"] = "1"
ns$propdmge[tolower(ns$propdmge) == ""] = "1"
ns$pdmg = ns$PROPDMG * as.numeric(ns$propdmge)

Then crop damage, too.


ns$cropdmge = ns$CROPDMGEXP
ns$cropdmge[tolower(ns$cropdmge) == "b"] = "1000000000"
ns$cropdmge[tolower(ns$cropdmge) == "m"] = "1000000"
ns$cropdmge[tolower(ns$cropdmge) == "k"] = "1000"
ns$cropdmge[tolower(ns$cropdmge) == "?"] = "1"
ns$cropdmge[tolower(ns$cropdmge) == ""] = "1"
ns$cdmg = ns$CROPDMG * as.numeric(ns$cropdmge)

Make EVTYPE as factor so aggregate function works.

ns$EVTYPE = as.factor(ns$EVTYPE)

We now process different impacts, such as casualty, injury and property/crop damage. First is social impacts, namely loss of life and injury.

casualty = aggregate(FATALITIES ~ EVTYPE, data = ns, sum)
inj = aggregate(INJURIES ~ EVTYPE, data = ns, sum)

Then the total ecomonical impacts for each type of event in the entire data.

ns$econdmg = ns$cdmg + ns$pdmg
econdmg = aggregate(econdmg ~ EVTYPE, data = ns, sum)

Results

Now let's visually examine our results. Those disasters could also cause casualties or injuries. While it is likely property and/or crop damages can be reduced by different approaches, such as insurance, hedge, preparation and etc, any loss of lives or health is unrecoverable. Therefore, it is the main focus of this analysis.

First, let's find which type of event caused most casualties.

casualty = casualty[order(-casualty$FATALITIES), ]
barplot(casualty$FATALITIES[1:5]/1000, names.arg = casualty$EVTYPE[1:5], cex.names = 0.7, 
    ylab = "Total casualties (thousand)", xlab = "Event type", main = "Five Types of Storms Events Resulting Most Casualties")

plot of chunk unnamed-chunk-7

We now see tornado caused more than five thousand deaths. How about the injuries? As shown in the plot below, tornado still caused most injuries and far more than any other storm events.

inj = inj[order(-inj$INJURIES), ]
barplot(inj$INJURIES[1:5]/1000, names.arg = inj$EVTYPE[1:5], cex.names = 0.7, 
    ylab = "Total injuries (thousand)", xlab = "Event type", main = "Five Types of Storms Events Resulting Most Injuries")

plot of chunk unnamed-chunk-8

Based on the above two figures, we conclude that tornadoes are the most hamful event to population health.

As tornado happens only in certain weather conditions and mostly in the so-called tornado valley, it is highly possible to predict such events and provide advanced warnings. By employing such prediction and alert procedure, loss of life and injury can be reduced.

Now let's check economical impacts. From the following bar chart, flood caused more than 140 billions of total property and crop damages in the data, followed by hurricane (70 billion), tornado (60 billion), storm surge (less than 50 billion) and hail (20 billion). Reducing flood requires great efforts, but the economical rewards can be significant.

econdmg = econdmg[order(-econdmg$econdmg), ]
barplot(econdmg$econdmg[1:5]/1e+09, names.arg = econdmg$EVTYPE[1:5], cex.names = 0.7, 
    ylab = "Total damage (billion dollars)", xlab = "Event type", main = "Five Types of Storms Events Resulting Most Damages")

plot of chunk unnamed-chunk-9

Copyright © 2016 thetazero.com All Rights Reserved. Privacy Policy