The effects of severe weather events on population health and economics between 1996 and 2011.

1. SYNOPSIS

The National Oceanic and Atmospheric Administration (NOAA) maintains a public database for storm events. The data that is collected here displays information on the type of storm and details like location, date, estimates for damage to property as well as the number of human victims of the storm. In this report we investigate which type of events are the most harmful to the population and economically between the period of 1996 and 2011.

From this analysis it is clear that tornadoes have the highest number of victims, causing the highest number of injuries and the second highest number of fatalities, closely following excessive heat events. In terms of economic costs however, weather types that occur on a larger scale and have a longer timespan than when compared to tornadoes, seem to have more impact. The leading weather type in terms of costs were floods, having the highest cost in property damage and coming in second in terms of crop damage, being preceded by droughts.

2. INTRODUCTION

Storms and other severe weather events can cause both public health and economic problems for communities and municipalities. Many severe events can result in fatalities, injuries, and property damage, and preventing such outcomes to the extent possible is a key concern.

This project involves exploring the U.S. National Oceanic and Atmospheric Administration’s (NOAA) storm database. This database tracks characteristics of major storms and weather events in the United States, including when and where they occur, as well as estimates of any fatalities, injuries, and property damage.

3. DATA

The data for this assignment come in the form of a comma-separated-value file compressed via the bzip2 algorithm to reduce its size. You can download the file from the course web site:

Storm Data [47Mb]

There is also some documentation of the database available. Here you will find how some of the variables are constructed/defined:

The events in the database start in the year 1950 and end in November 2011. In the earlier years of the database there are generally fewer events recorded, most likely due to a lack of good records. More recent years should be considered more complete.

4. DATA PROCESSING

Set options ans load libraries.

knitr::opts_chunk$set(echo = TRUE)

# Load in the needed libraries
library(ggplot2)

## Warning: package 'ggplot2' was built under R version 3.2.5

library(dplyr)

## Warning: package 'dplyr' was built under R version 3.2.5

## 
## Attaching package: 'dplyr'

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

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

library(plyr)

## Warning: package 'plyr' was built under R version 3.2.5

## -------------------------------------------------------------------------

## You have loaded plyr after dplyr - this is likely to cause problems.
## If you need functions from both plyr and dplyr, please load plyr first, then dplyr:
## library(plyr); library(dplyr)

## -------------------------------------------------------------------------

## 
## Attaching package: 'plyr'

## The following objects are masked from 'package:dplyr':
## 
##     arrange, count, desc, failwith, id, mutate, rename, summarise,
##     summarize

library(reshape2)

## Warning: package 'reshape2' was built under R version 3.2.5

# Adjust scientific notation
options(scipen=999)

Download, unzip and load data into a dataframe called storm. As data in the earlier years of the dataset is likely to be skewed due to a lack of good records, a subset of the more recent years was made. This date was arbitrarily set at 1996-01-01.

# Load in the data
setwd("/Users/sservaes/datasciencecoursera/RepData_PeerAssessment2")
if(!file.exists("StormData.csv.bz2")) {
        download.file("https://d396qusza40orc.cloudfront.net/repdata%2Fdata%2FStormData.csv.bz2", "StormData.csv.bz2")
}
storm <- read.csv("StormData.csv.bz2")

# Subset the dataset by removing the data before 1996-01-01
storm$BGN_DATE <- as.Date(storm$BGN_DATE, format = "%m/%d/%Y", "%Y")
storm_select <- storm[which(storm$BGN_DATE >= "1996-01-01"),]

In order to identify which severe weather type had the most influence on population health, the data was further processed as follows:

A new dataframe EVTYPE_pop was made to summarise variables according to EVTYPE.
The sum of FATALITIES was made according to EVTYPE in a variable fatalities.
The sum of INJURIES was made according to EVTYPE in a variable injuries.
FATALITIES and INJURIES were summed together according to EVTYPE in a variable affected_population.
In order to identify the worst severe weather types in respect to population health, a variable EVTYPE_pop_worst was made which grouped all the weather types that affected more than 1000 people (injuries or fatalities).

# Summarise the dataset by fatalities and injuries.
EVTYPE_pop <- ddply(storm_select, ~EVTYPE, summarise, 
                    fatalities = sum(FATALITIES), 
                    injuries = sum(INJURIES), 
                    affected_population = sum(FATALITIES)+sum(INJURIES))

# Determine the worst weather types (those that affect more that 1000 people).
EVTYPE_pop_worst <- EVTYPE_pop[which(EVTYPE_pop$affected_population>1000),]


# Melt the data according to EVTYPE by fatalities and injuries.
EVTYPE_pop_melt <- melt(EVTYPE_pop_worst, id.vars = "EVTYPE", 
                        measure.vars =  c("fatalities", "injuries"), 
                        variable.name="severity", value.name="health")

In order to identify which severe weather type had the most influence on economics, the data was further processed as follows:

Total damages were calculated taking the exponential into account.
In order to identify the worst severe weather types in respect to economics, a variable EVTYPE_eco_worst was made which grouped all the weather types that had a total cost higher than 5 billion USD (property or crops damage).

# Calculated total damage to property.
storm_select$PROPDMGTOT <- storm_select$PROPDMG
storm_select$PROPDMGTOT[which(storm_select$PROPDMGEXP=="K")] <- storm_select$PROPDMG[which(storm_select$PROPDMGEXP=="K")] * 1000
storm_select$PROPDMGTOT[which(storm_select$PROPDMGEXP=="M")] <- storm_select$PROPDMG[which(storm_select$PROPDMGEXP=="M")] * 1000000
storm_select$PROPDMGTOT[which(storm_select$PROPDMGEXP=="B")] <- storm_select$PROPDMG[which(storm_select$PROPDMGEXP=="B")] * 1000000000

# Calculated total damage to crops.
storm_select$CROPDMGTOT <- storm_select$CROPDMG
storm_select$CROPDMGTOT[which(storm_select$CROPDMGEXP=="K")] <- storm_select$CROPDMG[which(storm_select$CROPDMGEXP=="K")] * 1000
storm_select$CROPDMGTOT[which(storm_select$CROPDMGEXP=="M")] <- storm_select$CROPDMG[which(storm_select$CROPDMGEXP=="M")] * 1000000
storm_select$CROPDMGTOT[which(storm_select$CROPDMGEXP=="B")] <- storm_select$CROPDMG[which(storm_select$CROPDMGEXP=="B")] * 1000000000

# Summarise the dataset by property and crop damage.
EVTYPE_eco <- ddply(storm_select, ~EVTYPE, summarise, property_damage = sum(PROPDMGTOT), crop_damage = sum(CROPDMGTOT), total_costs = sum(PROPDMGTOT)+sum(CROPDMGTOT))

# Determine the worst weather types for economics (those that have a total cost higher than 5 billion USD).
EVTYPE_eco_worst <- EVTYPE_eco[which(EVTYPE_eco$total_costs>5000000000),]

# Melt the data according to EVTYPE by property_damage and crop_damage
EVTYPE_eco_melt <- melt(EVTYPE_eco_worst, id.vars = "EVTYPE", 
                        measure.vars =  c("crop_damage", "property_damage"), 
                        variable.name="damage_type", value.name="damage_cost")

5. RESULTS

5.1 Population Health

In order to identify which types of events (as indicated in the 𝙴𝚅𝚃𝚈𝙿𝙴 variable) are most harmful with respect to population health in the United States between 1996 and 2011, the variables fatalities and injuries were plotted between that time period.

# Data used for the plot.
print(EVTYPE_pop_worst, row.names = FALSE)

##             EVTYPE fatalities injuries affected_population
##     EXCESSIVE HEAT       1797     6391                8188
##        FLASH FLOOD        887     1674                2561
##              FLOOD        414     6758                7172
##               HEAT        237     1222                1459
##          HIGH WIND        235     1083                1318
##  HURRICANE/TYPHOON         64     1275                1339
##          LIGHTNING        651     4141                4792
##  THUNDERSTORM WIND        130     1400                1530
##            TORNADO       1511    20667               22178
##          TSTM WIND        241     3629                3870
##       WINTER STORM        191     1292                1483

# Generate the combined plot.
plot_pop <- ggplot(EVTYPE_pop_melt, aes(x = reorder(EVTYPE, -health), y = health/1000)) +
        geom_bar(stat = "identity", aes(fill = severity)) +
        theme(axis.text.x = element_text(angle = 45, size=8, hjust = 1, vjust = 1), 
              axis.title.x = element_blank()) + 
        ggtitle("Fatalities and Injuries (1996 - 2011)") + 
        ylab("Number of people (in thousands)") +
        scale_fill_manual(values=c("cornflowerblue", "coral1"), 
                          labels = c("Fatalities", "Injuries"), 
                          name = "Severity")
plot_pop

From the plots it is clear that in terms of fatalities the following 3 weather types have the highest numbers.

table_fatalities <- EVTYPE_pop[order(-EVTYPE_pop$fatalities),][1:3,1:2]
print(table_fatalities, row.names = FALSE)

##          EVTYPE fatalities
##  EXCESSIVE HEAT       1797
##         TORNADO       1511
##     FLASH FLOOD        887

The highest amount of injuries were seen in the following 3 weather types:

table_injuries <- EVTYPE_pop[order(-EVTYPE_pop$injuries),][1:3,c(1,3)]
print(table_injuries, row.names = FALSE)

##          EVTYPE injuries
##         TORNADO    20667
##           FLOOD     6758
##  EXCESSIVE HEAT     6391

The highest amount of injuries and fatalities combined were seen in the following 3 weather types:

table_combined <- EVTYPE_pop[order(-EVTYPE_pop$injuries),][1:3,c(1,4)]
print(table_combined, row.names = FALSE)

##          EVTYPE affected_population
##         TORNADO               22178
##           FLOOD                7172
##  EXCESSIVE HEAT                8188

From this analysis it is clear that between 1996 and 2011, weather type TORNADO affected the highest amount of people, with in total 22178 people affected. In terms of injuries weather type TORNADO had the highest amount of people with 20667 injuries. In terms of fatalities weather type EXCESSIVE HEAT had the highest amount of people with 1797 fatalities.

5.2 Economic consequences

# Data used for the plot.
EVTYPE_eco_worst <- EVTYPE_eco_worst[order(-EVTYPE_eco_worst$total_costs),]
print(EVTYPE_eco_worst, row.names = FALSE)

##             EVTYPE property_damage crop_damage  total_costs
##              FLOOD    143944833550  4974778400 148919611950
##  HURRICANE/TYPHOON     69305840000  2607872800  71913712800
##        STORM SURGE     43193536000        5000  43193541000
##            TORNADO     24616945710   283425010  24900370720
##               HAIL     14595143420  2476029450  17071172870
##        FLASH FLOOD     15222203910  1334901700  16557105610
##          HURRICANE     11812819010  2741410000  14554229010
##            DROUGHT      1046101000 13367566000  14413667000
##     TROPICAL STORM      7642475550   677711000   8320186550
##          HIGH WIND      5247860360   633561300   5881421660
##           WILDFIRE      4758667000   295472800   5054139800
##          TSTM WIND      4478026440   553915350   5031941790

# Generate the combined plot.
plot_eco <- ggplot(EVTYPE_eco_melt, aes(x = reorder(EVTYPE, -damage_cost), y = damage_cost/1000000000)) +
        geom_bar(stat = "identity", aes(fill = damage_type)) +
        theme(axis.text.x = element_text(angle = 45, size=8, hjust = 1, vjust = 1), axis.title.x = element_blank()) + 
        ggtitle("Economic Costs (1996 - 2011)") + 
        ylab("USD (in billions)") +
        scale_fill_manual(values=c("cornflowerblue", "coral1"), 
                          labels = c("Damage to crops", "Damage to property"), 
                          name = "Damage type")
plot_eco

From the plots it is clear that in terms of property damage the following 3 weather types have the highest numbers.

table_property <- EVTYPE_eco[order(-EVTYPE_eco$property_damage),][1:3,1:2]
print(table_property, row.names = FALSE)

##             EVTYPE property_damage
##              FLOOD    143944833550
##  HURRICANE/TYPHOON     69305840000
##        STORM SURGE     43193536000

The highest amount of crop damage were seen in the following 3 weather types:

table_crops <- EVTYPE_eco[order(-EVTYPE_eco$crop_damage),][1:3,c(1,3)]
print(table_crops, row.names = FALSE)

##     EVTYPE crop_damage
##    DROUGHT 13367566000
##      FLOOD  4974778400
##  HURRICANE  2741410000

The highest amount of total costs were seen in the following 3 weather types:

table_eco <- EVTYPE_eco[order(-EVTYPE_eco$total_costs),][1:3,c(1,4)]
print(table_eco, row.names = FALSE)

##             EVTYPE  total_costs
##              FLOOD 148919611950
##  HURRICANE/TYPHOON  71913712800
##        STORM SURGE  43193541000

From this analysis it is clear that between 1996 and 2011, weather type FLOOD had the highest amount of costs, with in total 148.92 billion USD. In terms of property damage weather type FLOOD had the highest cost with 143.94 billion USD. In terms of crop damage weather type DROUGHT had the highest cost with 13.37 billion USD.