Storms
Darius Kharazi
7/11/2018
Synopsis
Severe weather events can cause both economic and public health issues to those communities directly involved in the storm, and those that rely on their resources, as well. In recent years, there has been an increasing amount of storms that have destroyed cities and taken hundreds of lives. These storms will continue to grow in frequency and magnitude due to changes in the climate. Therefore, we should make a greater effort to better understand storms in order to minimize the amount of destruction caused by them through preparation.
The analysis will explore the Atmospheric Administration’s (NOAA) and the U.S. National Oceanic storm database to answer the given questions:
- Across the United States, which types of events (as indicated in the Event variable) are most harmful with respect to population health?
- Across the United States, which types of events have the greatest economic consequences?
After analyzing each type of storm, we’ll be able to prove that tornados are the most harmful in reference to public health, and floods are the most harmful in reference to the economy.
Import Statements
We’ve imported libraries that will help us read in the raw data and provide a means to publishing our results to RPubs. We’ve also imported libraries that help with creating transformations and reports about the raw data, specifically libraries that tidy, transform, and visualize the data. The “readr” package has been used to read in our raw data.
Data Processing
## Remove columns with NA
storms.df <- storms.df %>%
select(STATE, EVTYPE, FATALITIES, INJURIES, PROPDMG, PROPDMGEXP, CROPDMG, CROPDMGEXP)
## Rename columns
names(storms.df) <- c("State", "Type", "Deaths", "Injuries",
"P.Damage", "P.Units", "C.Damage", "C.Units")Some of the variables in the dataset are filled with NA values, which does not provide any useful information to us. Therefore, we should remove these columns from the dataset and only rename the useful variables to the following:
- State: The abbreviated name of the state where the storm was located.
- Type: The type of storm that occurred..
- Deaths: The number of fatalities caused by the storm.
- Injuries: The number of people injured in the storm.
- P.Damage: The property damage measured in dollars.
- P.Units: The units of property damage (B,M,K,H).
- C.Damage: The crop damage measured in dollars.
- C.Units: The units of crop damage (B,M,K,H).
Now that we have finished clening the variables, we should move on to cleaning the data.
## Redundant types of storms
head(sort(table(storms.df$Type), decreasing = TRUE), 50)##
## HAIL TSTM WIND THUNDERSTORM WIND
## 288661 219944 82563
## TORNADO FLASH FLOOD FLOOD
## 60652 54278 25326
## THUNDERSTORM WINDS HIGH WIND LIGHTNING
## 20843 20212 15755
## HEAVY SNOW HEAVY RAIN WINTER STORM
## 15708 11723 11433
## WINTER WEATHER FUNNEL CLOUD MARINE TSTM WIND
## 7026 6839 6175
## MARINE THUNDERSTORM WIND WATERSPOUT STRONG WIND
## 5812 3797 3566
## URBAN/SML STREAM FLD WILDFIRE BLIZZARD
## 3392 2761 2719
## DROUGHT ICE STORM EXCESSIVE HEAT
## 2488 2006 1678
## HIGH WINDS WILD/FOREST FIRE FROST/FREEZE
## 1533 1457 1342
## DENSE FOG WINTER WEATHER/MIX TSTM WIND/HAIL
## 1293 1104 1028
## EXTREME COLD/WIND CHILL HEAT HIGH SURF
## 1002 767 725
## TROPICAL STORM FLASH FLOODING EXTREME COLD
## 690 682 655
## COASTAL FLOOD LAKE-EFFECT SNOW FLOOD/FLASH FLOOD
## 651 636 624
## LANDSLIDE SNOW COLD/WIND CHILL
## 600 587 539
## FOG RIP CURRENT MARINE HAIL
## 538 470 442
## DUST STORM AVALANCHE WIND
## 427 386 341
## RIP CURRENTS STORM SURGE
## 304 261length(unique(storms.df$Type))## [1] 977If we examine the unique values for the types of storms in the data, we’re able to see that there are many redundant types of storms that are improperly formatted. Rather than having many messy entries without any real format, we would like to see grouped entries into a few types of storms, such as “tornado”, “flood”, etc.
## Replace the units for property damage
storms.df$P.Units <- recode(storms.df$P.Units, '0'=1, '-'=1, '?'=1, '+'=1,
'1'=10, '2'=10^2, 'h'=10^2, 'H'=10^2, '3'=10^3, 'K'=10^3, '4'=10^4,
'5'=10^5, '6'=10^6, 'm'=10^6, 'M'=10^6, '7'=10^7, '8'=10^8, 'B'=10^9)
## Replace the units for crop damage
storms.df$C.Units <- recode(storms.df$C.Units, '0'=1, '?'=1, '2'=10^2, 'k'=10^3, 'K'=10^3, 'm'=10^6, 'M'=10^6, 'B'=10^9)
## Replace values with "Avalanche"
storms.df$Type <- gsub(".*avalanche.*", "Avalanche", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*avalance.*", "Avalanche", storms.df$Type, ignore.case = TRUE)
## Replace values with "Landslide"
storms.df$Type <- gsub(".*landslide.*", "Landslide", storms.df$Type, ignore.case = TRUE)
## Replace values with "Thunder Storm"
storms.df$Type <- gsub(".*thunder.*", "Thunder Storm", storms.df$Type, ignore.case = TRUE)
## Replace values with "Tropical Storm"
storms.df$Type <- gsub(".*tropical.*", "Tropical Storm", storms.df$Type, ignore.case = TRUE)
## Replace values with "Mudslide"
storms.df$Type <- gsub(".*mud.*", "Mudslide", storms.df$Type, ignore.case = TRUE)
## Replace values with "Dust Storm"
storms.df$Type <- gsub(".*dust.*", "Dust Storm", storms.df$Type, ignore.case = TRUE)
## Replace values with "Volcano"
storms.df$Type <- gsub(".*volcan.*", "Volcano", storms.df$Type, ignore.case = TRUE)
## Replace values with "Waterspout"
storms.df$Type <- gsub(".*waterspout.*", "Waterspout", storms.df$Type, ignore.case = TRUE)
## Replace values with "Fog"
storms.df$Type <- gsub(".*fog.*", "Fog", storms.df$Type, ignore.case = TRUE)
## Replace values with "Rain"
storms.df$Type <- gsub(".*rain.*", "Heavy rain", storms.df$Type, ignore.case = TRUE)
## Replace values with "Lightning"
storms.df$Type <- gsub(".*lightning.*", "Lightning", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*lighting.*", "Lightning", storms.df$Type, ignore.case = TRUE)
## Replace values with "Tsunami"
storms.df$Type <- gsub(".*tsunami.*", "Tsunami", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*typhoon.*", "Tsunami", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*surge.*", "Tsunami", storms.df$Type, ignore.case = TRUE)
## Replace values with "Hurricane"
storms.df$Type <- gsub(".*hurricane.*", "Hurricane", storms.df$Type, ignore.case = TRUE)
## Replace values with "Heat"
storms.df$Type <- gsub(".*dry.*", "Heat", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*heat.*", "Heat", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*warm.*", "Heat", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*hot.*", "Heat", storms.df$Type, ignore.case = TRUE)
## Replace values with "Tornado"
storms.df$Type <- gsub(".*tornado.*", "Tornado", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*funnel.*", "Tornado", storms.df$Type, ignore.case = TRUE)
## Replace values with "Wind"
storms.df$Type <- gsub(".*wind.*", "Storm Winds", storms.df$Type, ignore.case = TRUE)
## Replace values with "Flood"
storms.df$Type <- gsub(".*flood.*", "Flood", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*fld.*", "Flood", storms.df$Type, ignore.case = TRUE)
## Replace values with "Winter Storm"
storms.df$Type <- gsub(".*winter.*", "Winter Storm", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*wintry.*", "Winter Storm", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*sleet.*", "Winter Storm", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*cold.*", "Winter Storm", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*hail.*", "Winter Storm", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*snow.*", "Winter Storm", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*blizzard.*", "Winter Storm", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*ice.*", "Winter Storm", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*icy.*", "Winter Storm", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*glaze.*", "Winter Storm", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*freez.*", "Winter Storm", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*frost.*", "Winter Storm", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*cool.*", "Winter Storm", storms.df$Type, ignore.case = TRUE)
## Replace values with "Wildfire"
storms.df$Type <- gsub(".*fire.*", "Wildfire", storms.df$Type, ignore.case = TRUE)
## Replace values with "Drought"
storms.df$Type <- gsub(".*drought.*", "Drought", storms.df$Type, ignore.case = TRUE)
## Replace values with "High Tide"
storms.df$Type <- gsub(".*tide.*", "High Tide", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*surf.*", "High Tide", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*current.*", "High Tide", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*seas.*", "High Tide", storms.df$Type, ignore.case = TRUE)
storms.df$Type <- gsub(".*waves.*", "High Tide", storms.df$Type, ignore.case = TRUE)Depending on the raw data, grouping messy data entries can require a great deal of manual string manipulation. Although we haven’t classified each and every storm type into its own distinct category of storm, we have significantly reduced the number of informative storm types. The majority of the remaining types of storms are not even insightful, since they are entered as a date that the incident occurred.
## Redundant types of storms
head(sort(table(storms.df$Type), decreasing = TRUE), 50)##
## Winter Storm Storm Winds Thunder Storm
## 334008 255382 109582
## Flood Tornado Lightning
## 86083 67673 15765
## Heavy rain Wildfire Waterspout
## 12238 4239 3860
## Heat Drought High Tide
## 3282 2488 2155
## Fog Tropical Storm Landslide
## 1883 757 613
## Dust Storm Tsunami Avalanche
## 589 530 388
## Hurricane OTHER Temperature record
## 200 48 43
## Mudslide MONTHLY PRECIPITATION MIXED PRECIPITATION
## 37 36 34
## Volcano SEICHE Record temperature
## 29 21 11
## SMOKE DENSE SMOKE MIXED PRECIP
## 11 10 10
## COASTAL STORM HEAVY MIX URBAN/SMALL STREAM
## 8 8 8
## LOW TEMPERATURE GUSTNADO HIGH WATER
## 7 6 6
## WET MICROBURST HIGH SWELLS MICROBURST
## 6 5 5
## RECORD HIGH RECORD TEMPERATURE ROTATING WALL CLOUD
## 5 5 5
## WALL CLOUD DAM BREAK Microburst
## 5 4 4
## MONTHLY TEMPERATURE Other RECORD LOW
## 4 4 4
## Wet Month Wet Year
## 4 4length(unique(storms.df$Type))## [1] 193We can see that the majority of the entries have been distinctly categorized, and a lot of the remaining storm types are summaries based on the date, rather than the storm. For future references, we could continue to reduce the number of storm types by categorizing each entry into an its distinct type of storm. For now, we will work with the current cleaned data, since we’ve categorized the majority of storms.
Results
## Find the storms affecting the most people
most.harmful <- storms.df %>%
mutate(Total = Injuries + Deaths) %>%
group_by(Type) %>%
summarize(Injury = sum(Injuries), Death = sum(Deaths), Total = sum(Total)) %>%
arrange(desc(Total)) %>%
top_n(n = 10, wt = Total) %>%
gather(Harm, Casualties, 2:3)
## Plot storms causing the most harm
ggplot(most.harmful, aes(x = reorder(Type, Total), y = Casualties, fill = Harm)) +
geom_bar(stat = "identity") +
labs(x = "Storm Type", y = "Casualties", fill = "Type") +
coord_flip()
As we can see, tornados injure and kill the most people in the data. There are nearly 90,000 injuries caused by tornados, whereas heat causes the second most injuries (9,000), which is an enormous difference. Tornados are also the most significant killer out of any other storm, but it shouldbe noted that a significant percentage of the fatalities caused by storms is due to heat-related incidents (3,000).
## Find the storms affecting economies the most
most.cost <- storms.df %>%
group_by(Type) %>%
summarize(Crop = sum(C.Damage * C.Units, na.rm = TRUE),
Property = sum(P.Damage * P.Units, na.rm = TRUE)) %>%
mutate(Total = Crop + Property) %>%
arrange(desc(Total)) %>%
top_n(n = 10, wt = Total) %>%
gather(CostType, Cost, 2:3)
## Plot storms causing the most harm
ggplot(most.cost, aes(x = reorder(Type, Total), y = Cost, fill = CostType)) +
geom_bar(stat = "identity") +
labs(x = "Storm Type", y = "Cost", fill = "Type") +
coord_flip()
When examining the economic costs caused by the different types of storms, we can see that floods create the the largest percentage of property costs out of any other storm. Floods are are the most common and costliest natural disaster, since they create nearly 168 billion dollars in property costs and 12 billion dollars in crop costs. The second most costly storm is the tsunami, which creates nearly 112 billion dollars in property costs and 2 billion dollars in crop costs. Clearly, floods are significantly more costly overall than any other storm, but it should be noted that droughts are the most costly natural disaster in terms of crop costs, since they create 13 billion dollars in crop costs.