Effects of severe weather events in the USA 1950

Synopsis

In this report we want to analyse how severe weather events affet the population health and his economic effects. For that we will explore the NOAA Storm Database. The events in the database start in the year 1950 and end in November 2011.

We want to address the following questions:
- Across the United States, which types of events (as indicated in the EVTYPE variable) are most harmful with respect to population health (fatalities and injuries)?
- Across the United States, which types of events have the greatest economic consequences?

Data Processing

Loading packages

Here are listed the packages needed for the analysis

library(dplyr)
library(reshape2)
library(ggplot2)

Loading data

Data can be downloaded from here

DF <- read.csv("repdata-data-StormData.csv.bz2")

After reading the file we check the dimmension of file and inspect the first columns

dim(DF)

## [1] 902297     37

head(DF[,1:12])

##   STATE__           BGN_DATE BGN_TIME TIME_ZONE COUNTY COUNTYNAME STATE
## 1       1  4/18/1950 0:00:00     0130       CST     97     MOBILE    AL
## 2       1  4/18/1950 0:00:00     0145       CST      3    BALDWIN    AL
## 3       1  2/20/1951 0:00:00     1600       CST     57    FAYETTE    AL
## 4       1   6/8/1951 0:00:00     0900       CST     89    MADISON    AL
## 5       1 11/15/1951 0:00:00     1500       CST     43    CULLMAN    AL
## 6       1 11/15/1951 0:00:00     2000       CST     77 LAUDERDALE    AL
##    EVTYPE BGN_RANGE BGN_AZI BGN_LOCATI END_DATE
## 1 TORNADO         0                            
## 2 TORNADO         0                            
## 3 TORNADO         0                            
## 4 TORNADO         0                            
## 5 TORNADO         0                            
## 6 TORNADO         0

Cleaning Data

We will extract the needed values for the analysis

DF_clean <- DF[, c("BGN_DATE","EVTYPE", "FATALITIES", "INJURIES","PROPDMG", "PROPDMGEXP", "CROPDMG", "CROPDMGEXP")]

DF_clean <- DF_clean[complete.cases(DF_clean),] # We remove rows with NAs

There are different nammes for the same event, so we will consolidate some of the most common to make it more consistent

DF_clean$EVTYPE[grepl("HEAT",      DF_clean$EVTYPE)]<-"HEAT"
DF_clean$EVTYPE[grepl("WARM",      DF_clean$EVTYPE)]<-"HEAT"
DF_clean$EVTYPE[grepl("RIP CURRENT",   DF_clean$EVTYPE)]<-"RIP CURRENT"
DF_clean$EVTYPE[grepl("FLOOD",     DF_clean$EVTYPE)]<-"FLOOD"
DF_clean$EVTYPE[grepl("THUNDERSTORM",      DF_clean$EVTYPE)]<-"THUNDERSTORM"
DF_clean$EVTYPE[grepl("WIND",      DF_clean$EVTYPE)]<-"WIND"
DF_clean$EVTYPE[grepl("SNOW",      DF_clean$EVTYPE)]<-"SNOW"
DF_clean$EVTYPE[grepl("FIRE",      DF_clean$EVTYPE)]<-"WILDFIRE"
DF_clean$EVTYPE[grepl("HURRICANE", DF_clean$EVTYPE)]<-"HURRICANE"
DF_clean$EVTYPE[grepl("TYPHOON",   DF_clean$EVTYPE)]<-"HURRICANE"

colnames(DF_clean)[2] <- "events" #Rename the value "EVTYPE"" by "events"

Next we clean the Property Damage Data (see documentation here)

I would assume that the letters correspond to exponents, example “B” (Billion) would mean 10^9
NA will be transformed to 0 -> 10^0 = 1

levels(DF_clean$PROPDMGEXP)<-c(levels(DF_clean$PROPDMGEXP), 0:9)

DF_clean$PROPDMGEXP[DF_clean$PROPDMGEXP=="B"]<-"9"
DF_clean$PROPDMGEXP[DF_clean$PROPDMGEXP=="M"]<-"6"
DF_clean$PROPDMGEXP[DF_clean$PROPDMGEXP=="m"]<-"6"
DF_clean$PROPDMGEXP[DF_clean$PROPDMGEXP=="K"]<-"3"
DF_clean$PROPDMGEXP[DF_clean$PROPDMGEXP=="H"]<-"2"
DF_clean$PROPDMGEXP[DF_clean$PROPDMGEXP=="h"]<-"2"
DF_clean$PROPDMGEXP[DF_clean$PROPDMGEXP==" "]<-"0"
DF_clean$PROPDMGEXP[DF_clean$PROPDMGEXP==""]<-"0"
DF_clean$PROPDMGEXP[DF_clean$PROPDMGEXP=="+"]<-"0"
DF_clean$PROPDMGEXP[DF_clean$PROPDMGEXP=="-"]<-"0"
DF_clean$PROPDMGEXP[DF_clean$PROPDMGEXP=="?"]<-"0"
DF_clean$PROPDMGEXP[is.na(DF_clean$PROPDMGEXP)]<-"0"
DF_clean$PROPDMGEXP <- as.numeric(as.character(DF_clean$PROPDMGEXP))

table(DF_clean$PROPDMGEXP)

## 
##      0      1      2      3      4      5      6      7      8      9 
## 466164     25     20 424669      4     28  11341      5      1     40

Now we clean up exponential of Crop Damage variable We clean the values folowing the same rules as with Property Damage, example “B” (Billion) would mean 10^9

levels(DF_clean$CROPDMGEXP)<-c(levels(DF_clean$CROPDMGEXP), 0:9)

DF_clean$CROPDMGEXP[DF_clean$CROPDMGEXP=="B"]<-"9"
DF_clean$CROPDMGEXP[DF_clean$CROPDMGEXP=="M"]<-"6"
DF_clean$CROPDMGEXP[DF_clean$CROPDMGEXP=="m"]<-"6"
DF_clean$CROPDMGEXP[DF_clean$CROPDMGEXP=="K"]<-"3"
DF_clean$CROPDMGEXP[DF_clean$CROPDMGEXP=="k"]<-"3"
DF_clean$CROPDMGEXP[DF_clean$CROPDMGEXP=="?"]<-"0"
DF_clean$CROPDMGEXP[DF_clean$CROPDMGEXP==""]<-"0"
DF_clean$CROPDMGEXP[is.na(DF_clean$CROPDMGEXP)]<-"0"
DF_clean$CROPDMGEXP <- as.numeric(as.character(DF_clean$CROPDMGEXP))

table(DF_clean$CROPDMGEXP)

## 
##      0      2      3      6      9 
## 618439      1 281853   1995      9

Now we create a new variable with the total damage This results by adding Property and Crop damages and evaluating the exponential (cleaned before)

DF_clean$Total_DM <- ((DF_clean$PROPDMG*10^(DF_clean$PROPDMGEXP)) + 
                              (DF_clean$CROPDMG*10^(DF_clean$CROPDMGEXP)))

Results

Events related to fatalities

To answer the question of which types of events are most harmful with respect to population health, we will count the amount of fatalities per event.

fatalities_per_event <- DF_clean %>% 
        select(events, FATALITIES) %>%
        group_by(events) %>% 
        summarise(fatalities=sum(FATALITIES)) %>%
        mutate (percent_fatalities=(100*fatalities)/sum(fatalities)) %>%
        arrange(desc(fatalities))

fatalities_per_event$percent_fatalities<-
        format(round(fatalities_per_event$percent_fatalities, 2), nsmall = 2)

head(fatalities_per_event,10)

## Source: local data frame [10 x 3]
## 
##          events fatalities percent_fatalities
## 1       TORNADO       5633              37.19
## 2          HEAT       3178              20.98
## 3         FLOOD       1523              10.06
## 4          WIND       1237               8.17
## 5     LIGHTNING        816               5.39
## 6   RIP CURRENT        577               3.81
## 7     AVALANCHE        224               1.48
## 8  THUNDERSTORM        210               1.39
## 9  WINTER STORM        206               1.36
## 10         SNOW        162               1.07

From the data we can derive that:
- Tornados, Heat and Floods are the main causes of fatalities (68%)
- The top 10 events cover more than 90% of all fatalities

Events related to injuries

To answer the question of which types of events are most harmful with respect to population health, we will count the amount of injuries per event.

injuries_per_event <- DF_clean %>% 
        select(events, INJURIES) %>%
        group_by(events) %>% 
        summarise(injuries=sum(INJURIES)) %>%
        mutate (percent_injuries=(100*injuries)/sum(injuries)) %>%
        arrange(desc(injuries))

injuries_per_event$percent_injuries<- 
        format(round(injuries_per_event$percent_injuries, 2), nsmall = 2)

head(injuries_per_event,10)

## Source: local data frame [10 x 3]
## 
##          events injuries percent_injuries
## 1       TORNADO    91346            65.00
## 2          HEAT     9173             6.53
## 3          WIND     9056             6.44
## 4         FLOOD     8603             6.12
## 5     LIGHTNING     5230             3.72
## 6  THUNDERSTORM     2479             1.76
## 7     ICE STORM     1975             1.41
## 8      WILDFIRE     1608             1.14
## 9          HAIL     1361             0.97
## 10    HURRICANE     1331             0.95

From the data we can derive that:
- Tornados are the main cause of injuries (65%)
- The rest of injuries correspond to a long list of events with percentages lower than 10%

Now we will plot major events and the amount of injuries and fatalities

DF_merge<-merge(fatalities_per_event, injuries_per_event, by="events")

# We reshape the Data Frame to get a long formmat. This helps with the plotting
DF_merge_long<-DF_merge %>%
        select(., events, fatalities, injuries) %>%
        group_by(events) %>%
        reshape2::melt(., id=c("events")) %>%
        arrange(., desc(value))

names(DF_merge_long) <- c("Events", "Type", "Amount")

ggplot(head(DF_merge_long,11), aes(x=Events, y=Amount)) +
        geom_bar(stat="identity", aes(fill = Events == "TORNADO" )) +
        scale_fill_manual(values = c("black", "red") ) +
        facet_wrap(~ Type) +
        ggtitle("Population health damage by type of weather event") + 
        theme(plot.title = element_text(lineheight=.8, face="bold")) +
        theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
        theme(legend.position="none")

In the plot we see again that:
- Tornado is the most harmful event, followed by Heat and Flood

Events related to economic consequences

To calculate the economic damage we will put together property and crop damage.

DF_Total_DM<-DF_clean %>%
        select(., events, Total_DM) %>%
        group_by(events) %>%
        summarise(damage=sum(Total_DM)) %>%
        mutate (percent_damage=(100*damage)/sum(damage)) %>%
        arrange(., desc(damage))

head(DF_Total_DM, 10)

## Source: local data frame [10 x 3]
## 
##            events       damage percent_damage
## 1           FLOOD 180413708935      37.796506
## 2       HURRICANE  90762527810      19.014666
## 3         TORNADO  57362333947      12.017356
## 4     STORM SURGE  43323541000       9.076242
## 5            HAIL  18761221986       3.930459
## 6         DROUGHT  15018672000       3.146398
## 7            WIND  13922997768       2.916855
## 8       ICE STORM   8967041360       1.878587
## 9        WILDFIRE   8904910130       1.865570
## 10 TROPICAL STORM   8382236550       1.756071

ggplot(head(DF_Total_DM,11), aes(x=events, y=(damage/(10^9)))) +
        geom_bar(stat="identity", aes(fill = events == "FLOOD" )) +
        scale_fill_manual(values = c("black", "red") ) +
        ggtitle("Economical damage by type of weather event") + 
        labs(x="Events", y = "Damages (US$ billions)") +
        theme(plot.title = element_text(lineheight=.8, face="bold")) +
        theme(axis.text.x = element_text(angle = 45, hjust = 1)) +
        theme(legend.position="none")

In the plot we see that:
- Flood causes the biggest damage (37%), followed by Hurricane, Tornado and Storm (77%)

Summary of results

Now we collect all the results in one list:

Tornados, Heat and Floods are the main causes of fatalities (68%)
Tornados are the main cause of injuries (65%)
Flood causes the biggest damage (37%), followed by Hurricane, Tornado and Storm (77%)

Effects of severe weather events in the USA 1950 - 2011

Javier Sáenz

Oktober 2015