Analysis of Severe Weather Events

Synopsis

The purpose of this analysis is to examine severe weather events as well as consequences related to such events. Specifically, this analysis addresses two main questions:

What events result in the most harmful losses to general population?
What are the biggest economic losses caused by severe weather conditions?

To answer these questions, we use data from the U.S. National Oceanic and Atmospheric Administration (NOAA). NOAA tracks major storms and weather events in the United States, including dates, types, as well as estimates of any fatalities, injuries, and property damage.

Data Processing

First, the data were downloaded from NOAA website and loaded into R. Given the size of the dataset, we used readr package, which provides a faster way to read tabular data. We also used use dplyr and tidyr package to manipulate the data. ggplot2 was used to visualize the results.

library(readr)
library(dplyr)
library(tidyr)
library(ggplot2)

storm <- read_csv('repdata-data-StormData.csv.bz2')

## 
|================================================================================| 100%  535 MB

storm$BGN_DATE <- as.POSIXct(strptime(storm$BGN_DATE, "%m/%d/%Y %H:%M:%S"))
dim(storm)

## [1] 902297     37

min(storm$BGN_DATE);max(storm$BGN_DATE)

## [1] "1950-01-03 EST"

## [1] "2011-11-30 EST"

As it can be seen, there are over 900,000 records and 37 variables in the file. The data set includes almost 62 years of data.

Results

This section presents key findings for the two main questions. Please note that while individual estimates can vary by year depending on the severity of weather conditions, this analysis provides findings across all years combined.

1. Across the United States, which types of events are most harmful with respect to population health?

Let’s look at top five most deadly events as well as those resulting in a biggest number of injuries. We grouped by event type and summed up number of fatalities and injuries. The table below summarizes results.

p <- storm %>% 
  group_by(EVTYPE) %>%
  summarise(Deaths=sum(FATALITIES),Injuries=sum(INJURIES)) %>% 
  top_n(5) %>% 
  arrange(desc(Deaths))

## Selecting by Injuries

knitr::kable(p, caption="Table 1: Most Harmful Weather Events 1950-2011")

Table 1: Most Harmful Weather Events 1950-2011
EVTYPE	Deaths	Injuries
TORNADO	5633	91346
EXCESSIVE HEAT	1903	6525
LIGHTNING	816	5230
TSTM WIND	504	6957
FLOOD	470	6789

As it can be seen, tornado are the most deadly type of event resulting in a largest number of deaths and injuries. The chart below shows this information graphically.

p %>% gather(Type,N, 2:3) %>% 
  ggplot(aes(x=EVTYPE,y=N))+geom_bar(stat='identity')+facet_wrap(~Type,scales = "free_y")+
  labs(title='Figure 1: Most Harmful Weather Events 1950-2011')

2. Across the United States, which types of events have the greatest economic consequences?

To measure economic economic losses, we created a new variable called total_dam. This variable is a sum of two variables - property damages and crop damages. As before, we examined top five events with greatest economic losses. Variable PRODMEXP contains units ( billions, millions, etc.) in which losses are measured. (CROPDMGEXP also contains units but it is empty). Before performing the analysis, some recoding was be done to ensure consistency.

f<-function(x){switch(as.character(x), H=100, K=1000, M=1000000,B=1000000000,1)}
storm$PROPDMGEXP<- sapply(storm$PROPDMGEXP, f)

storm <- storm %>% 
  filter(PROPDMG > 0) %>% 
  mutate(total_dam = PROPDMG*PROPDMGEXP+CROPDMG)

p <- storm %>% 
  group_by(EVTYPE) %>% 
  summarise(Losses=sum(total_dam)) %>% 
  top_n(5) %>% 
  arrange(desc(Losses))

## Selecting by Losses

p$Losses <- prettyNum(p$Losses,big.mark = ",")

knitr::kable(p,caption="Table 2: Most Economically Devastating Weather Events 1950-2011")

Table 2: Most Economically Devastating Weather Events 1950-2011
EVTYPE	Losses
FLOOD	144,657,858,149
HURRICANE/TYPHOON	69,305,844,748
TORNADO	56,925,751,854
STORM SURGE	43,323,536,005
FLASH FLOOD	16,140,985,677