PeerAssesment2.Rmd

---
title: "Analysis of NOAA Storm Database"
output: html_document
---

# Synopsis

This assignment analyzes the US NOAA Storm Database to address the following two questions:

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

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

### Environment
- Windows 8.1 64 bit
- R v3.1.1
- Rstudio v0.98.1062.0


# Data Processing

```{r message=FALSE}
library(stringr)
library(plyr)
library(dplyr)
library(tidyr)
library(ggplot2)
```

```{r cache=TRUE}
fileurl <- "https://d396qusza40orc.cloudfront.net/repdata%2Fdata%2FStormData.csv.bz2"
fileurl <- sub("^https", "http", fileurl)

filename <- "StormData.csv.bz2"
if (!file.exists(filename)) {download.file(fileurl, filename)}

stormdata <- read.csv(bzfile(filename), stringsAsFactors=FALSE)
dim(stormdata)
```

```{r }
stormdf <- tbl_df(stormdata) %>%
    filter(FATALITIES > 0 | INJURIES > 0 | PROPDMG > 0 | CROPDMG > 0) %>%
    mutate(YEAR = as.integer(str_extract(BGN_DATE, "\\d{4}"))) %>%
    filter(YEAR >= 1996) %>%
    select(EVTYPE, FATALITIES, INJURIES, PROPDMG, PROPDMGEXP, CROPDMG, CROPDMGEXP)

length(unique(stormdf$EVTYPE))
```

```{r }
clean_evtype <- function(x) {
    x <- toupper(str_trim(x))
    
    if (grepl("ACCIDENT|DROWNING", x)) "OTHER"
    else if (grepl("HIGH TIDE", x)) "COASTAL FLOOD"
    else if (grepl("COASTAL ?STORM", x)) "STORM SURGE/TIDE"
    else if (grepl("COASTAL|CSTL|BEACH", x)) "COASTAL FLOOD"
    else if (grepl("EXTREME", x)) "EXTREME COLD/WIND CHILL"
    else if (grepl("COLD", x)) "COLD/WIND CHILL"
    else if (grepl("FREEZING", x)) "FREEZING FOG"
    else if (grepl("FOG", x)) "DENSE FOG"
    else if (grepl("BLOWING DUST", x)) "DUST STORM"
    else if (grepl("^HEAT$", x)) "HEAT"
    else if (grepl("HEAT", x)) "EXCESSIVE HEAT"
    else if (grepl("THERMIA|WARM", x)) "HEAT"
    else if (grepl("FLASH|DAM BREAK|HIGH WATER", x)) "FLASH FLOOD"
    else if (grepl("LAKESHORE FLOOD", x)) "LAKESHORE FLOOD"
    else if (grepl("FLOOD|FLD", x)) "FLOOD"
    else if (grepl("MARINE T[^ ]*M\\b", x)) "MARINE THUNDERSTORM WIND"
    else if (grepl("MARINE", x)) x
    else if (grepl("ICE STORM", x)) "ICE STORM"
    else if (grepl("FROST|FREEZE|ICE|ICY", x)) "FROST/FREEZE"
    else if (grepl("LAKE.*SNOW", x)) "LAKE-EFFECT SNOW"
    else if (grepl("SNOW", x)) "HEAVY SNOW"
    else if (grepl("RAIN", x)) "HEAVY RAIN"
    else if (grepl("HAIL", x)) "HAIL"
    else if (grepl("SURF", x)) "HIGH SURF"
    else if (grepl("WINTER STORM", x)) "WINTER STORM"
    else if (grepl("WINTER|WINTRY|GLAZE|PRECIP", x)) "WINTER WEATHER"
    else if (grepl("SLIDE|SLUMP", x)) "DEBRIS FLOW"
    else if (grepl("RIVER", x)) "FLOOD"
    else if (grepl("GUSTY", x)) "STRONG WIND"
    else if (grepl("NON.?TSTM", x)) "OTHER"
    else if (grepl("THUNDERSTORM|TSTM|BURST", x)) "THUNDERSTORM WIND"
    else if (grepl("STRONG WIND", x)) "STRONG WIND"
    else if (grepl("WIND", x)) "HIGH WIND"
    else if (grepl("HURRICANE|TYPHOON", x)) "HURRICANE(TYPHOON)"
    else if (grepl("RIP", x)) "RIP CURRENT"
    else if (grepl("TORNADO|LANDSPOUT", x)) "TORNADO"
    else if (grepl("TROPICAL", x)) x
    else if (grepl("STORM|TIDE", x)) "STORM SURGE/TIDE"
    else if (grepl("FIRE", x)) "WILDFIRE"
    else if (grepl("WAVE|SWELL| SEAS$", x)) "HIGH SURF"
    else x
}
```

```{r }
stormdf$EVTYPE <- sapply(stormdf$EVTYPE, clean_evtype)

length(unique(stormdf$EVTYPE))
```

```{r }
symbol2value <- function(x) {
    if (x == "K") 1e3
    else if (x == "M") 1e6
    else if (x == "B") 1e9
    else 1
}
```

```{r }
stormdf$PROPDMG <- stormdf$PROPDMG * sapply(stormdf$PROPDMGEXP, symbol2value)
stormdf$CROPDMG <- stormdf$CROPDMG * sapply(stormdf$CROPDMGEXP, symbol2value)

stormdf <- stormdf %>%
    select(-PROPDMGEXP, -CROPDMGEXP) %>%
    group_by(EVTYPE) %>%
    summarize(FATALITIES = sum(FATALITIES), INJURIES = sum(INJURIES), 
              PROPDMG = sum(PROPDMG), CROPDMG = sum(CROPDMG)) %>%
    mutate(TOTHEALTHDMG = FATALITIES + INJURIES, TOTECONOMICDMG = PROPDMG + CROPDMG)
```


# Results

### Types of events that are most harmful with respect to population health:
```{r }
top_health_dmg <- stormdf %>%
    select(-c(PROPDMG, CROPDMG, TOTECONOMICDMG)) %>%
    filter(FATALITIES >= 275 | INJURIES >= 875) %>%
    arrange(desc(FATALITIES), desc(TOTHEALTHDMG))

top_health_dmg

fatalities <- select(top_health_dmg, EVTYPE, FATALITIES)
colnames(fatalities) <- c("EVTYPE", "TOTHEALTHDMG")

injuries <- select(top_health_dmg, EVTYPE, INJURIES)
colnames(injuries) <- c("EVTYPE", "TOTHEALTHDMG")

# health_dmg <- rbind(fatalities, injuries)
health_dmg = rbind(mutate(fatalities, factor = rep("FATALITIES",12)),
                   mutate(injuries, factor = rep("INJURIES",12)))

ggplot(health_dmg, aes(x = EVTYPE, y = TOTHEALTHDMG, fill = factor)) +
    geom_bar(stat = "identity") +
    labs(title = "Event Types that have Most Health Impact",
         x = "Event Type", y = "Total Health Damage") +
    theme(axis.text.x = element_text(angle = 30, hjust = 1))
```

### Types of events that have the greatest economic consequences:
```{r }
top_economic_dmg <- stormdf %>%
    select(-c(FATALITIES, INJURIES, TOTHEALTHDMG)) %>%
    mutate(TOTECONOMICDMG = TOTECONOMICDMG / 1e9) %>% # amount to billion
    filter(TOTECONOMICDMG >= 2) %>%
    arrange(desc(TOTECONOMICDMG))

top_economic_dmg

ggplot(top_economic_dmg, aes(x = EVTYPE, y = TOTECONOMICDMG)) +
    geom_bar(stat = "identity", fill = "#7700FF") +
    labs(title = "Event Types that have Most Economic Impact",
         x = "Event Type", y = "Total Economic Damage (in Billion US$)") +
    theme(axis.text.x = element_text(angle = 30, hjust = 1))
```

### Note
This analysis was done under Peer Assesment 2 of Coursera course *Reproducible Research*.