In this report we analyse the public health and economic impact of storms and other major weather events in United States. Many of such severe events can result in fatalities, injuries, crop and property damage, and preventing such outcomes is a key concern for federal and state government and for municipalities.
We explored the U.S. National Oceanic and Atmospheric Administration's (NOAA) storm database and, using the data from 2001 to 2011, we found which type of event had most health and economic impact in those 10 years span.
This report could be of interest to government or municipal managers who might be responsible for preparing for severe weather events and will need to prioritize resources for different types of events.
Note: this report was made as an assignment for the Reproducible Research Coursera course. The report is made with RMarkdown and Knitr. The source code is available on github
The U.S. National Oceanic and Atmospheric Administration's (NOAA) storm database tracks characteristics of major storms and weather events in the United States, including when and where they occur, as well as estimates of any fatalities, injuries, and property damage. The events in the database start in the year 1950 and end in November 2011.
We obtain a compressed file from the Coursera course site: Storm data [47Mb], and we read the raw text file included.
storm_data_raw <- read.csv(bzfile("StormData.csv.bz2"))
We check the dimension and the first few rows of the data set.
storm_data_dim <- dim(storm_data_raw)
storm_data_dim
## [1] 902297 37
head(storm_data_raw)
## STATE__ BGN_DATE BGN_TIME TIME_ZONE COUNTY COUNTYNAME STATE
## 1 1 4/18/1950 0:00:00 0130 CST 97 MOBILE AL
## 2 1 4/18/1950 0:00:00 0145 CST 3 BALDWIN AL
## 3 1 2/20/1951 0:00:00 1600 CST 57 FAYETTE AL
## 4 1 6/8/1951 0:00:00 0900 CST 89 MADISON AL
## 5 1 11/15/1951 0:00:00 1500 CST 43 CULLMAN AL
## 6 1 11/15/1951 0:00:00 2000 CST 77 LAUDERDALE AL
## EVTYPE BGN_RANGE BGN_AZI BGN_LOCATI END_DATE END_TIME COUNTY_END
## 1 TORNADO 0 0
## 2 TORNADO 0 0
## 3 TORNADO 0 0
## 4 TORNADO 0 0
## 5 TORNADO 0 0
## 6 TORNADO 0 0
## COUNTYENDN END_RANGE END_AZI END_LOCATI LENGTH WIDTH F MAG FATALITIES
## 1 NA 0 14.0 100 3 0 0
## 2 NA 0 2.0 150 2 0 0
## 3 NA 0 0.1 123 2 0 0
## 4 NA 0 0.0 100 2 0 0
## 5 NA 0 0.0 150 2 0 0
## 6 NA 0 1.5 177 2 0 0
## INJURIES PROPDMG PROPDMGEXP CROPDMG CROPDMGEXP WFO STATEOFFIC ZONENAMES
## 1 15 25.0 K 0
## 2 0 2.5 K 0
## 3 2 25.0 K 0
## 4 2 2.5 K 0
## 5 2 2.5 K 0
## 6 6 2.5 K 0
## LATITUDE LONGITUDE LATITUDE_E LONGITUDE_ REMARKS REFNUM
## 1 3040 8812 3051 8806 1
## 2 3042 8755 0 0 2
## 3 3340 8742 0 0 3
## 4 3458 8626 0 0 4
## 5 3412 8642 0 0 5
## 6 3450 8748 0 0 6
The data set contains 902297 events and 37 variables.
We are interested in the following variables.
Begin date of the event. The events could span many days, but for our purposes we date them with the begin date. We extract the year and we will use it to filter the events registered between 2001 and 2011.
There are two reasons for this filter:
The NOAA storm database code book reports 48 event type. The event types in the data set are more than 9 hundred.
length(unique(storm_data_raw$EVTYPE))
## [1] 985
Some difference are caused by upper and lower cases, but many of them are caused by incorrect imputation, especially in the early years. There is not an easy way to correct them so we looked for event types with big values for fatalities and damage and use regular expression substitution to correct them.
storm_data_corrected <- storm_data_raw
storm_data_corrected$EVTYPE <- toupper(storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^(SMALL )?HAIL.*", "HAIL", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("TSTM|THUNDERSTORMS?", "THUNDERSTORM", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("STORMS?", "STORM", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("WINDS?|WINDS?/HAIL", "WIND", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("RAINS?", "RAIN", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^TH?UN?DEE?RS?TO?RO?M ?WIND.*|^(SEVERE )?THUNDERSTORM$|^WIND STORM$|^(DRY )?MI[CR][CR]OBURST.*|^THUNDERSTORMW$", "THUNDERSTORM WIND", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^COASTAL ?STORM$|^MARINE ACCIDENT$", "MARINE THUNDERSTORM WIND", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^FLOODS?.*|^URBAN/SML STREAM FLD$|^(RIVER|TIDAL|MAJOR|URBAN|MINOR|ICE JAM|RIVER AND STREAM|URBAN/SMALL STREAM)? FLOOD(ING)?S?$|^HIGH WATER$|^URBAN AND SMALL STREAM FLOODIN$|^DROWNING$|^DAM BREAK$", "FLOOD", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^FLASH FLOOD.*|^RAPIDLY RISING WATER$", "FLASH FLOOD", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("WATERSPOUTS?", "WATERSPOUT", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("WEATHER/MIX", "WEATHER", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("CURRENTS?", "CURRENT", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^WINDCHILL$|^COLD.*|^LOW TEMPERATURE$|^UNSEASONABLY COLD$", "COLD/WIND CHILL", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^EXTREME WIND ?CHILL$|^(EXTENDED|EXTREME|RECORD)? COLDS?$", "EXTREME COLD/WIND CHILL", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^WILD/FOREST FIRE$|^(WILD|BRUSH|FOREST)? ?FIRES?$", "WILDFIRE", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^RAIN/SNOW$|^(BLOWING|HEAVY|EXCESSIVE|BLOWING|ICE AND|RECORD)? ?SNOWS?.*", "HEAVY SNOW", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^FOG$", "DENSE FOG", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^(GUSTY|NON-SEVERE|NON ?-?THUNDERSTORM)? ?WIND.*|^ICE/STRONG WIND$", "STRONG WIND", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("SURGE$", "SURGE/TIDE", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("CLOUDS?", "CLOUD", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^FROST[/\\]FREEZE$|^FROST$|^(DAMAGING)? ?FREEZE$|^HYP[OE]R?THERMIA.*|^ICE$|^(ICY|ICE) ROADS$|^BLACK ICE$|^ICE ON ROAD$", "FROST/FREEZE", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^GLAZE.*|^FREEZING (RAIN|DRIZZLE|RAIN/SNOW|SPRAY$)$|^WINTRY MIX$|^MIXED PRECIP(ITATION)?$|^WINTER WEATHER MIX$|^LIGHT SNOW$|^FALLING SNOW/ICE$|^SLEET.*", "SLEET", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^HURRICANE.*", "HURRICANE/TYPHOON", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^HEAT WAVES?$|^UNSEASONABLY WARM$|^WARM WEATHER$", "HEAT", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^(EXTREME|RECORD/EXCESSIVE|RECORD) HEAT$", "EXCESSIVE HEAT", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^HEAVY SURF(/HIGH SURF)?.*$|^(ROUGH|HEAVY) SEAS?.*|^(ROUGH|ROGUE|HAZARDOUS) SURF.*|^HIGH WIND AND SEAS$|^HIGH SURF.*", "HIGH SURF", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^LAND(SLUMP|SLIDE)?S?$|^MUD ?SLIDES?$|^AVALANCH?E$", "AVALANCHE", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^UNSEASONABLY WARM AND DRY$|^DROUGHT.*|^HEAT WAVE DROUGHT$", "DROUGHT", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^TORNADO.*", "TORNADO", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^TROPICAL STORM.*", "TROPICAL STORM", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^MARINE MISHAP$|^HIGH WIND/SEAS$", "MARINE HIGH WIND", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^HIGH WIND.*", "HIGH WIND", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^HIGH SEAS$", "MARINE STRONG WIND", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^RIP CURRENT.*", "RIP CURRENT", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^WATERSPOUT.*", "WATERSPOUT", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^EXCESSIVE RAINFALL$|^RAIN.*|^TORRENTIAL RAINFALL$|^(HEAVY|HVY)? (RAIN|MIX|PRECIPITATION).*", "HEAVY RAIN", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^FOG.*", "FREEZING FOG", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^WINTER STORM.*", "WINTER STORM", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^THUNDERSNOW$|^ICE STORM.*", "ICE STORM", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("WAVES?|SWELLS?", "SURF", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^LIGHTNING.*", "LIGHTNING", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^WHIRLWIND$|^GUSTNADO$|^TORNDAO$", "TORNADO", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^COASTAL FLOOD.*", "COASTAL FLOOD", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^TYPHOON", "HURRICANE/TYPHOON", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^EROSION/CSTL FLOOD$|^COASTAL FLOOD/EROSION$|^COASTAL SURGE/TIDE$", "COASTAL FLOOD", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^ASTRONOMICAL HIGH TIDE$", "STORM SURGE/TIDE", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^(GROUND)? ?BLIZZARD.*$", "BLIZZARD", storm_data_corrected$EVTYPE)
storm_data_corrected$EVTYPE <- gsub("^DUST STORM.*$", "DUST STORM", storm_data_corrected$EVTYPE)
Fatalities and injuries estimated for the event. These values are used to estimate the public health impact for type of events.
Property and crop damage estimated for the event. These values are used to estimate the economic impact for type of events.
These variables are associated with PROPDMGEXP and CROPDMGEXP which are used as exponents to interpret the numeric values for the damage. There are not much information in the data code book about these variables.
The only symbols with a clear meaning are:
We sum the damage values grouped by the symbols.
property_data_exp <- storm_data_raw %.%
group_by(PROPDMGEXP) %.%
summarise(property_damage_per_exp = sum(PROPDMG)) %.%
arrange(PROPDMGEXP)
print(xtable(property_data_exp), type="html")
| PROPDMGEXP | property_damage_per_exp | |
|---|---|---|
| 1 | 527.41 | |
| 2 | - | 15.00 |
| 3 | ? | 0.00 |
| 4 | + | 117.00 |
| 5 | 0 | 7108.30 |
| 6 | 1 | 0.00 |
| 7 | 2 | 12.00 |
| 8 | 3 | 20.00 |
| 9 | 4 | 14.50 |
| 10 | 5 | 210.50 |
| 11 | 6 | 65.00 |
| 12 | 7 | 82.00 |
| 13 | 8 | 0.00 |
| 14 | B | 275.85 |
| 15 | h | 2.00 |
| 16 | H | 25.00 |
| 17 | K | 10735292.10 |
| 18 | m | 38.90 |
| 19 | M | 140694.45 |
crop_data_exp <- storm_data_raw %.%
group_by(CROPDMGEXP) %.%
summarise(crop_damage_per_exp = sum(CROPDMG)) %.%
arrange(CROPDMGEXP)
print(xtable(crop_data_exp), type="html")
| CROPDMGEXP | crop_damage_per_exp | |
|---|---|---|
| 1 | 11.00 | |
| 2 | ? | 0.00 |
| 3 | 0 | 260.00 |
| 4 | 2 | 0.00 |
| 5 | B | 13.61 |
| 6 | k | 436.00 |
| 7 | K | 1342955.91 |
| 8 | m | 10.00 |
| 9 | M | 34140.80 |
We see that the symbols without a clear meaning are associated with minimal values. So we use only the H, K, M, B symbols to interpret the damage amounts and we clear the amounts for the other symbols.
We create a function to decode the symbols and return a multiplier for the amounts. We use as base value the million of dollars, as we are interested only in the most impactful damage. This function will be used to prepare the final data set.
# Function to decode the EXP symbol
decode_exp <- function(exp_symbol) {
# Normalize to millions of dollars
if (toupper(exp_symbol) == "B") exp <-1000
else if (toupper(exp_symbol) == "M") exp <-1
else if (toupper(exp_symbol) == "K") exp <- 1/1000
else if (toupper(exp_symbol) == "H") exp <- 1/10000
# Don't know how to interpet other values
else exp <-0
return(exp)
}
decode_exp_v <- Vectorize(decode_exp)
To add up and compare amounts of different years, we have to convert them in constant dollars, i.e. inflation adjusted amounts.
To compute a conversion factor for each year, we download the Consumer Price Index for All Urban Consumers: All Items from Federal Reserve Economic Data, we average the monthly CPI to obtain an annual value, and we compute a factor based at 2011 that could be multiplied with the amounts to adjust for the inflation.
# Get Consumer Price Index from Federal Reserve Economic Data
getSymbols("CPIAUCSL", src='FRED')
# CPI is monthly. Calculate an annual average.
annual_cpi <- apply.yearly(CPIAUCSL, mean)
# Calculate conversion factor using 2011 as the base year
conversion_factor <- 1 / annual_cpi * as.numeric(annual_cpi['2011'])
The final data set, used for the analysis, is obtained by:
storm_data <- storm_data_corrected %.%
filter(year(mdy_hms(BGN_DATE)) >= 2001) %.%
mutate(property_damage_exp = decode_exp_v(PROPDMGEXP)
,crop_damage_exp = decode_exp_v(CROPDMGEXP)
,event_year = year(mdy_hms(BGN_DATE))
,factor_cd = as.numeric(conversion_factor[as.character(event_year)])
,property_damage = PROPDMG * property_damage_exp * factor_cd
,crop_damage = CROPDMG * crop_damage_exp * factor_cd) %.%
select(event_year
,event_type = EVTYPE
,fatalities = FATALITIES
,injuries = INJURIES
,property_damage
,crop_damage)
storm_data_final_dim <- dim(storm_data)
storm_data_final_dim
[1] 488692 6
print(xtable(head(storm_data), digits = c(0,0,0,0,0,6,6)), type = "html")
| event_year | event_type | fatalities | injuries | property_damage | crop_damage | |
|---|---|---|---|---|---|---|
| 1 | 2001 | THUNDERSTORM WIND | 0 | 0 | 0.002541 | 0.000000 |
| 2 | 2001 | THUNDERSTORM WIND | 0 | 0 | 0.019058 | 0.000000 |
| 3 | 2001 | HAIL | 0 | 0 | 0.000000 | 0.000000 |
| 4 | 2001 | THUNDERSTORM WIND | 0 | 0 | 0.006353 | 0.000000 |
| 5 | 2001 | THUNDERSTORM WIND | 0 | 0 | 0.006353 | 0.000000 |
| 6 | 2001 | THUNDERSTORM WIND | 0 | 0 | 0.003812 | 0.000000 |
The final data set contains 488692 events and 6 variables.
The analysis aims to answer two question:
We consider as “harmful to population health”“ the events with registered fatalities and injuries.
We select only the events with fatalities and injuries.
fatalities_registered <- storm_data[storm_data$fatalities != 0, 3]
injuries_registered <- storm_data[storm_data$injuries != 0, 4]
fatalities_registered_sum <- sum(fatalities_registered)
injuries_registered_sum <- sum(injuries_registered)
summary(fatalities_registered)
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 1.00 1.00 1.00 1.71 1.00 158.00
summary(injuries_registered)
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 1.0 1.0 1.0 5.7 3.0 1150.0
We can see that in the 2001-2011 period were registered 5517 fatalities and 32330 injuries.
The distribution of fatalities has a median of 1, a third quartile of 1 and a max value of 158. This mean that more than the 75% of the events have 1 fatalities and there are few events with a large number of fatalities.
Injuries has a similar distribution with a median of 1, a third quartile of 3 and a max value of 1150.
We can visually confirm these distribution with a boxplot of fatalities and injuries (figure 1). As the distribution is highly skewed towards 1, we have plotted the y axis as the logarithm in base 10 of fatalities and injuries to make the distribution clearer.
par(mfrow=c(1,2), mar=c(5,2,2,2), oma=c(0,0,3,0))
boxplot(fatalities_registered, log = "y", xlab = "Fatalities", ylim = c(1,1200))
boxplot(injuries_registered, log = "y", xlab = "Injuries", ylim = c(1,1200))
mtext("Fig. 1 - Fatalities and Injuries (log10)", side=3, line=1, outer=TRUE)
We find out the top 10 event types by fatalities and we compute the average number of fatalities per event types. For all the event type the average is less than 1.
event_type_fatalities_summary <- storm_data %.%
group_by(event_type) %.%
summarise(fatalities_per_type = sum(fatalities), event_count = n()) %.%
mutate(average_fatalities = fatalities_per_type / event_count) %.%
arrange(desc(fatalities_per_type, event_count))
print(xtable(head(event_type_fatalities_summary, n = 10), digits = c(0,0,0,0,2)), type="html")
| event_type | fatalities_per_type | event_count | average_fatalities | |
|---|---|---|---|---|
| 1 | TORNADO | 1152 | 16520 | 0.07 |
| 2 | EXCESSIVE HEAT | 856 | 1059 | 0.81 |
| 3 | FLASH FLOOD | 573 | 38412 | 0.01 |
| 4 | RIP CURRENT | 433 | 570 | 0.76 |
| 5 | LIGHTNING | 414 | 8779 | 0.05 |
| 6 | FLOOD | 270 | 19939 | 0.01 |
| 7 | HEAT | 230 | 735 | 0.31 |
| 8 | THUNDERSTORM WIND | 227 | 154645 | 0.00 |
| 9 | AVALANCHE | 200 | 895 | 0.22 |
| 10 | EXTREME COLD/WIND CHILL | 143 | 1152 | 0.12 |
We find out the top 10 event types by injuries and we compute the average number of injuries per event types. Here stand out HURRICAN/TYPHOON with 9 average injuries per event, and EXCESSIVE HEAT with 3.
event_type_injuries_summary <- storm_data %.%
group_by(event_type) %.%
summarise(injuries_per_type = sum(injuries), event_count = n()) %.%
mutate(average_injuries = injuries_per_type / event_count) %.%
arrange(desc(injuries_per_type, event_count))
print(xtable(head(event_type_injuries_summary, n = 10), digits = c(0,0,0,0,2)), type="html")
| event_type | injuries_per_type | event_count | average_injuries | |
|---|---|---|---|---|
| 1 | TORNADO | 14331 | 16520 | 0.87 |
| 2 | EXCESSIVE HEAT | 3242 | 1059 | 3.06 |
| 3 | THUNDERSTORM WIND | 2913 | 154645 | 0.02 |
| 4 | LIGHTNING | 2622 | 8779 | 0.30 |
| 5 | HURRICANE/TYPHOON | 1291 | 133 | 9.71 |
| 6 | HEAT | 1222 | 735 | 1.66 |
| 7 | WILDFIRE | 1099 | 3294 | 0.33 |
| 8 | FLASH FLOOD | 780 | 38412 | 0.02 |
| 9 | HIGH WIND | 557 | 15569 | 0.04 |
| 10 | HAIL | 488 | 154472 | 0.00 |
Merging the two list we can see that the event types with the big impact on public health are: TORNADO, EXCESSIVE HEAT, FLASH FLOOD, LIGHTNING, HEAT, THUNDERSTORM WIND.
We consider as "with economic consequences”“ the events with registered property and crop damage.
We select only the events with property and crop damage.
property_damage_registered <- storm_data[storm_data$property_damage != 0, 5]
crop_damage_registered <- storm_data[storm_data$crop_damage != 0, 6]
property_damage_registered_sum <- sum(property_damage_registered)
crop_damage_registered_sum <- sum(crop_damage_registered)
summary(property_damage_registered)
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0 0 0 3 0 128000
summary(crop_damage_registered)
## Min. 1st Qu. Median Mean 3rd Qu. Max.
## 0.0 0.0 0.0 1.9 0.1 1740.0
We can see that in the 2001-2011 period were registered 366608 property damage and 22989 crop damage (millions of dollars).
The distribution of property damage has a median of 0, a third quartile of 0 and a max value of 128000 (million of dollars). This mean that more than the 75% of the events have sub-million property damage and there are few events with damage in the millions and even in the billions of dollars.
Crop damage has a similar distribution with a median of 0, a third quartile of 0.1 and a max value of 1740 (millions of dollars).
We can visually confirm these distribution with a boxplot of property and crop damage (figure 2). As the distribution is highly skewed towards 0, we have plotted the y axis as the logarithm in base 10 of property and crop damage to make the distribution clearer.
par(mfrow=c(1,2), mar=c(5,2,2,2), oma=c(0,0,3,0))
boxplot(property_damage_registered, log = "y", xlab = "Property damage", ylim = c(.00001,100000))
boxplot(crop_damage_registered, log = "y", xlab = "Crop damage", ylim = c(.00001,100000))
mtext("Fig. 2 - Property and Crop damage - Millions of dollars (log10)", side=3, line=1, outer=TRUE)
We find out the top 10 event types by property damage and we compute the average number of property damage per event types. Here stand out HURRICAN/TYPHOON with 631 million of dollars average property damage per event, and STORM SURGE/TIDE with 147 million of dollars.
event_type_property_damage_summary <- storm_data %.%
group_by(event_type) %.%
summarise(property_damage_per_type = sum(property_damage), event_count = n()) %.%
mutate(average_property_damage = property_damage_per_type / event_count) %.%
arrange(desc(property_damage_per_type, event_count))
print(xtable(head(event_type_property_damage_summary, n = 10), digits = c(0,0,0,0,2)), type="html")
| event_type | property_damage_per_type | event_count | average_property_damage | |
|---|---|---|---|---|
| 1 | FLOOD | 148450 | 19939 | 7.45 |
| 2 | HURRICANE/TYPHOON | 83926 | 133 | 631.02 |
| 3 | STORM SURGE/TIDE | 54595 | 371 | 147.16 |
| 4 | TORNADO | 20202 | 16520 | 1.22 |
| 5 | HAIL | 12917 | 154472 | 0.08 |
| 6 | FLASH FLOOD | 12711 | 38412 | 0.33 |
| 7 | TROPICAL STORM | 8965 | 605 | 14.82 |
| 8 | THUNDERSTORM WIND | 5834 | 154645 | 0.04 |
| 9 | WILDFIRE | 5722 | 3294 | 1.74 |
| 10 | HIGH WIND | 5668 | 15569 | 0.36 |
We find out the top 10 event types by crop damage and we compute the average number of crop damage per event types. Here stands out HURRICAN/TYPHOON with 26 million of dollars average crop damage per event.
event_type_crop_damage_summary <- storm_data %.%
group_by(event_type) %.%
summarise(crop_damage_per_type = sum(crop_damage), event_count = n()) %.%
mutate(average_crop_damage = crop_damage_per_type / event_count) %.%
arrange(desc(crop_damage_per_type, event_count))
print(xtable(head(event_type_crop_damage_summary, n = 10), digits = c(0,0,0,0,2)), type="html")
| event_type | crop_damage_per_type | event_count | average_crop_damage | |
|---|---|---|---|---|
| 1 | DROUGHT | 7862 | 1933 | 4.07 |
| 2 | FLOOD | 3879 | 19939 | 0.19 |
| 3 | HURRICANE/TYPHOON | 3552 | 133 | 26.71 |
| 4 | HAIL | 1880 | 154472 | 0.01 |
| 5 | FROST/FREEZE | 1232 | 1407 | 0.88 |
| 6 | FLASH FLOOD | 888 | 38412 | 0.02 |
| 7 | THUNDERSTORM WIND | 712 | 154645 | 0.00 |
| 8 | HIGH WIND | 578 | 15569 | 0.04 |
| 9 | EXCESSIVE HEAT | 550 | 1059 | 0.52 |
| 10 | HEAVY RAIN | 465 | 9102 | 0.05 |
Merging the two list we can see that the event types with the big impact on public health are: FLOOD, HURRICANE/TYPHOON, HAIL, FLASH FLOOD, THUNDERSTORM WIND, HIGH WIND.