Synopsis

The basic goal of this assignment is to explore the NOAA Storm Database and answer the following questions:

Across the United States, which types of events are most harmful with respect to population health?
Across the United States, which types of events have the greatest economic consequences?

Data Processing

First of all, we install and load the needed packages

library(data.table)
library(stringdist)
library(lubridate)

## 
## Attaching package: 'lubridate'

## The following objects are masked from 'package:data.table':
## 
##     hour, isoweek, mday, minute, month, quarter, second, wday,
##     week, yday, year

## The following object is masked from 'package:base':
## 
##     date

library(dplyr)

## 
## Attaching package: 'dplyr'

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

## The following objects are masked from 'package:data.table':
## 
##     between, first, last

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

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

library(gghighlight)

## Loading required package: ggplot2

library(gridExtra)

## 
## Attaching package: 'gridExtra'

## The following object is masked from 'package:dplyr':
## 
##     combine

Next, we load the storm data, and we lighten the data by keeping only the needed variables

storm_data <- fread("repdata_data_StormData.csv.bz2", header = T, sep = ',')
filtered_storm_data <- storm_data[, c("EVTYPE", "BGN_DATE", "BGN_TIME", "TIME_ZONE", "END_DATE", "END_TIME", "FATALITIES", "INJURIES", "CROPDMG", "CROPDMGEXP", "PROPDMG", "PROPDMGEXP")]

Putting the exponential of prop and crop damages values to a better format

filtered_storm_data$propmultiplication<-1
filtered_storm_data$propmultiplication[filtered_storm_data$PROPDMGEXP=="H"]<-100      
filtered_storm_data$propmultiplication[filtered_storm_data$PROPDMGEXP=="K"]<-1000      
filtered_storm_data$propmultiplication[filtered_storm_data$PROPDMGEXP=="M"]<-1000000   
filtered_storm_data$propmultiplication[filtered_storm_data$PROPDMGEXP=="B"]<-1000000000

filtered_storm_data$cropmultiplication<-1
filtered_storm_data$cropmultiplication[filtered_storm_data$CROPDMGEXP=="H"]<-100       
filtered_storm_data$cropmultiplication[filtered_storm_data$CROPDMGEXP=="K"]<-1000      
filtered_storm_data$cropmultiplication[filtered_storm_data$CROPDMGEXP=="M"]<-1000000   
filtered_storm_data$cropmultiplication[filtered_storm_data$CROPDMGEXP=="B"]<-1000000000 
# use the new factors to calculate two new columns with the property and crop cost
filtered_storm_data$propTotal <- filtered_storm_data$PROPDMG * filtered_storm_data$propmultiplication
filtered_storm_data$cropTotal <- filtered_storm_data$CROPDMG * filtered_storm_data$cropmultiplication

The event types are not well represented in the data, many events are present more than once with slightly different names, the function amatch will help put the data in a better format

official_event_types <- read.csv("event_types.csv")
i <- amatch(filtered_storm_data$EVTYPE, official_event_types$Event.Name,maxDist = 50)
filtered_storm_data$EVTYPE <- official_event_types$Event.Name[i]

Now we create a year column

filtered_storm_data$BGN_DATE <- as.Date(filtered_storm_data$BGN_DATE, "%m/%d/%Y %H:%M:%S")
filtered_storm_data$year <- year(filtered_storm_data$BGN_DATE)

Plots

We want to have the mean of all damage numbers for each event type by year. After a little bit of exploration, we find that records before 1992 are not enough. So we keep only data recorded after 1992

global_mean <- filtered_storm_data %>% group_by(EVTYPE, year) %>%
    summarise(fatalities.mean = mean(FATALITIES), injuries.mean = mean(INJURIES),
              prop.mean = mean(propTotal), crop.mean = mean(cropTotal))

#we take only years after 1992, because there is not enough records before
global_mean <- global_mean[global_mean$year>=1992,]

Ploting the mean of fatalities and injuries by year and we highlights the events corresponding to the highest values

plot1 <- ggplot(global_mean) +
  geom_line(aes(year, fatalities.mean, colour = EVTYPE)) +
        xlab("Year") + ylab("Fatalities") +
        ggtitle("Fatalities per event per year")+
        coord_cartesian(xlim=c(1992, 2011), ylim = c(0,5)) +
        gghighlight(max(fatalities.mean)>1.5) +
        theme_minimal()

## label_key: EVTYPE

plot2 <- ggplot(global_mean) +
  geom_line(aes(year, injuries.mean, colour = EVTYPE)) +
        xlab("Year") + ylab("Injuries") +
        ggtitle("Injuries per event type per year")+
        coord_cartesian(xlim=c(1992, 2011), ylim = c(0,20)) +
        gghighlight(max(injuries.mean)>8) +
        theme_minimal()

## label_key: EVTYPE

grid.arrange(plot1, plot2, nrow = 2)

Doing the same for prop and crop damages values

plot3 <- ggplot(global_mean) +
  geom_line(aes(year, prop.mean, colour = EVTYPE)) +
        xlab("Year") + ylab("Prop damages") +
        ggtitle("Prop damages per event per year")+
        coord_cartesian(xlim=c(1992, 2011), ylim = c(0,2e+08)) +
        gghighlight(max(prop.mean)>0.5e+08) +
        theme_minimal()

## label_key: EVTYPE

plot4 <- ggplot(global_mean) +
  geom_line(aes(year, crop.mean, colour = EVTYPE)) +
        xlab("Year") + ylab("Crop damages") +
        ggtitle("Crop damages per event type per year")+
        coord_cartesian(xlim=c(1992, 2011), ylim = c(0,2e+07)) +
        gghighlight(max(crop.mean)>1e+07) +
        theme_minimal()

## label_key: EVTYPE

grid.arrange(plot3, plot4, nrow = 2)

Results

The first figure shows that Excessive Heat and Seiche are the events causing the highest fatality and injury numbers, they are the most harmful to health based on the analysis.

The second figure shows that Dense Smoke and Avalanche are the events causing the highest prop damages. For Crop damages, Ice storms and Dense Fog tend to be the most important causes.

Analysis of Public health and Economy damages caused by weather events

Synopsis

Data Processing

Plots

Results