This analysis is based on data from the National Oceanic and Atmospheric Administration (NOAA) gathered data about storms over the last ~65 years. The database tracks multiple kinds of storms, their location, type, damage records, and other ancillary factors. For the history of tracked data, impact to population health was greatest among Tornadoes, which accounted for a large share compared to other events of both fatalities and injuries. Additionally, Tornadoes are the major event driving economic impacts, principally through property damage. Though less total economic damage, hail accounts for the largest portion of crop damage among other analyzed factors.
Raw Data Link: https://d396qusza40orc.cloudfront.net/repdata%2Fdata%2FStormData.csv.bz2
Load the dependent libraries.
library(ggplot2)
library(reshape2)
Load the unzipped data into a dataframe, and convert dates to date object format. Add an additional column for the year extracted from the date.
storm <- read.csv(file = bzfile("repdata-data-StormData.csv.bz2"), header = TRUE,
na.strings = "NA")
storm$BGN_DATE <- as.Date(storm$BGN_DATE, "%m/%d/%Y")
storm$END_DATE <- as.Date(storm$END_DATE, "%m/%d/%Y")
storm$YEAR <- as.POSIXlt(storm$BGN_DATE)$year + 1900
The data contains records for a very large array of Event Types.
length(unique(storm$EVTYPE))
## [1] 985
To analyze those events with the greatest impact, we can restrict analysis by those events the occur the most frequently and have caused the greatest damage. To accomplish this, a table of the occurence frequency of events is calculated, and ordered as a data frame, restricting to a vector of the 8 most common occurring events to be used later as a filter.
freq <- as.data.frame(table(storm$EVTYPE))
names(freq) <- c("EVTYPE", "COUNT")
freq <- freq[order(freq$COUNT, decreasing = TRUE), ]
freq_sel <- freq[1:8, "EVTYPE"]
freq_sel
## [1] HAIL TSTM WIND THUNDERSTORM WIND
## [4] TORNADO FLASH FLOOD FLOOD
## [7] THUNDERSTORM WINDS HIGH WIND
## 985 Levels: HIGH SURF ADVISORY COASTAL FLOOD ... WND
To understand effect to population health, analysis will concentrate on the impact of Injuries and Fatalities attributable to each event.
For the subject events, aggregate Injuries and Fatalities and merge into a common spline (to ensure all events are included in resulting frame regardless of aggregation inclusion). Then the data is melted to put all measures in a common vector to more easily feed the graphics.
agg_event <- data.frame(EVTYPE = freq_sel)
agg_event <- merge(agg_event, (aggregate(INJURIES ~ EVTYPE, data = storm, FUN = "sum")))
agg_event <- merge(agg_event, (aggregate(FATALITIES ~ EVTYPE, data = storm,
FUN = "sum")))
melt_event <- melt(data = agg_event, id.vars = c("EVTYPE"))
Plot Bar chart of the data to show both total impact as well as proportion by sub-factors.
ggplot(data = melt_event, aes(x = EVTYPE)) + geom_bar(aes(y = value, fill = variable),
stat = "identity") + ggtitle("Total Injuries and Fatalities by Event Type") +
xlab("Event") + ylab("Number of Injuries/Fatalities")
In terms of both factors representing population health, Tornadoes relate to the major share of injuries and fatalities. Tornadoes are not only the largest event driving negative effects to population health, they account for a larger share than the other 7 major factors combined. Additionally, as would be colloquially expected, there is a far larger impact in terms of numbers of injuries than fatalities. The relative value of them is not a subject of comparison of this analysis, nor definiable by the scope of the data present. However, give that both factor categories are at their highest levels attributable to Tornadoes, the outcome of the claim is equivalent nevertheless.
As sentinals for economic effect, the monetary value of property damage and crop damage was aggreagated as attributible for each event, and then quantified as an overall impact.
The initial transformation and aggregation of the data uses a similar apprach as the previous question for economic factors.
agg_econ <- data.frame(EVTYPE = freq_sel)
agg_econ <- merge(agg_econ, (aggregate(PROPDMG ~ EVTYPE, data = storm, FUN = "sum")))
agg_econ <- merge(agg_econ, (aggregate(CROPDMG ~ EVTYPE, data = storm, FUN = "sum")))
names(agg_econ) <- c("EVTYPE", "PropertyDamage", "CropDamage")
melt_econ <- melt(data = agg_econ, id.vars = c("EVTYPE"))
The data is plotted with a bar chart to show scale and comparable effects.
ggplot(data = melt_econ, aes(x = EVTYPE)) + geom_bar(aes(y = value, fill = variable),
stat = "identity") + ggtitle("Total Economic Damages by Event Type") + xlab("Event") +
ylab("Value of Economic Damage")
In terms of economic damage, Tornadoes once again account for the largest impact among the other key factors. Among key factors, the ratio of value of crop damages to property damages is very small with the exception of hail. With regard to hail, it's ratio of crop damage to property damage is nearly equivalent.