===========================================================================================
===========================================================================================
This study aims to analyze the consequences of weather events as recorded by U.S. National Oceanic and Atmospheric Administration's (NOAA) from 1950 to 2011. The study will point out the weather events causing the highest damage to human properties and agriculture crops and the events causing the highest number of injuries and fatalities.
Injuries, fatalities, property and crop damages will be analyzed separately because it is author's belief that, for instance, injuries and fatalities cannot be summed for their different nature and meaning; additionally damages to properties may be of higher interest than damages to crops, or viceversa, depending on the use made of this document and its audience.
First aggregated data will be shown. Afterwards, results will be split by decades pointing out differences, which are probably due to different methods of gathering data over time. In the end, a comparison between events from 90s and events from 00s will be reported. It will be shown that while tornados have been the most harmful and floods have been the most destructive weather events taking the time period as a whole, if we only consider the last 2 decades, hot weather have become more relevant than tornados as major cause of fatalities while floods keeps on being the worst enemy for properties.
A code-visible data processing section is included for reader to be able to reproduce the study and validate it.
=========================================================================================== In this following section, it will be described the manipulation used to clean, reshape, classify and plot the data.
===========================================================================================
## Please install and load the R.utils in order to 'bunzip' the .bz2 file
## Make also sure to have either the StormData.csv.bz2 or StormData.csv
## stored in your working directory
## Unzip and load the source file
if ((!file.exists("StormData.csv")) & (!file.exists("StormData.csv.bz2"))) {
message("please make sure to download either the file StormData.csv.bz2, or StormData.csv in your working directory")
return()
}
if ((!file.exists("StormData.csv")) & (file.exists("StormData.csv.bz2"))) {
bunzip2("StormData.csv.bz2", "StormData.csv")
}
raw_data <- read.csv("StormData.csv")
## Subsets over the columns of interest, eliminating the rest in order to
## make the huge dataframe easier to handle
data <- raw_data[, c("STATE", "BGN_DATE", "EVTYPE", "FATALITIES", "INJURIES",
"PROPDMG", "PROPDMGEXP", "CROPDMG", "CROPDMGEXP")]
## Un-factiring the char variables
data$PROPDMGEXP <- as.character(data$PROPDMGEXP)
data$CROPDMGEXP <- as.character(data$CROPDMGEXP)
data$EVTYPE <- as.character(data$EVTYPE)
## Let's take a first look to the data frame
str(data)## 'data.frame': 902297 obs. of 9 variables:
## $ STATE : Factor w/ 72 levels "AK","AL","AM",..: 2 2 2 2 2 2 2 2 2 2 ...
## $ BGN_DATE : Factor w/ 16335 levels "1/1/1966 0:00:00",..: 6523 6523 4242 11116 2224 2224 2260 383 3980 3980 ...
## $ EVTYPE : chr "TORNADO" "TORNADO" "TORNADO" "TORNADO" ...
## $ FATALITIES: num 0 0 0 0 0 0 0 0 1 0 ...
## $ INJURIES : num 15 0 2 2 2 6 1 0 14 0 ...
## $ PROPDMG : num 25 2.5 25 2.5 2.5 2.5 2.5 2.5 25 25 ...
## $ PROPDMGEXP: chr "K" "K" "K" "K" ...
## $ CROPDMG : num 0 0 0 0 0 0 0 0 0 0 ...
## $ CROPDMGEXP: chr "" "" "" "" ...
===========================================================================================
The general idea of what we are going to implement in the next is to quantify the damages, injuries and fatalities and sum these quantities grouping by event type. To do so, first of all we need to translate damages from string to numerical values expressed in a uniform scale. In other words, we first need to translate string multipliers (data$PROPDMGEXP and data$CROPDMGEXP columns for properties and crop damages respectively) into numeric multipliers and then multiply these multipliers with the corresponding damage values, i.e. data$PROPDMG and CROPDMG columns for properties and crop damages respectively. Once calculated the Prop and Crop damages of each atmospheric event, we will then sum up these values grouping by event type (column data$EVTYPE).
This is what we are going to do in the next lines. Let's start with a quick look and to the data and let's clean it.
##
## + - 0 1 2 3 4 5 6
## 465934 5 1 216 25 13 4 4 28 4
## 7 8 ? B H K M h m
## 5 1 8 40 6 424665 11330 1 7
##
## 0 2 ? B K M k m
## 618413 19 1 7 9 281832 1994 21 1
As you can see, from the tables above, the exponential multipliers CROPDMGEXP and PROPDMGEXP are characters that we first need to clean from errors and then translate into numerical factors.
Values as 1, 2, 3 have to be translated into exponential multiplying factors 10^1, 10^2, 10^3
Chars like B(or b), M (or m), K (or k), H (or h) have to be translated into 10^9, 10^6, 10^3, 10^2 respectively.
Conversely, blank values or characters like "+", "-", ? etc are typing errors and will be translated into a multiplier equal to 1.
Let's start by replacing errors.
## Let's first build the conversion table
fact_as_char <- c("B", "b", "M", "m", "K", "k", "h", "H", "9", "8", "7", "6",
"5", "4", "3", "2", "1", "0")
fact_as_num <- c(10^9, 10^9, 10^6, 10^6, 10^3, 10^3, 10^2, 10^2, 10^9, 10^8,
10^7, 10^6, 10^5, 10^4, 10^3, 10^2, 10^1, 1)
factors <- data.frame(fact_as_char, fact_as_num)
## Those EXP string in our dataframe which are not included in the above
## conversion dataframe, are considered errors and will be replaced with
## string '0' which in turn (see the table above) corresponds to the numeric
## value of 1.
data[!(data$PROPDMGEXP %in% fact_as_char), ]$PROPDMGEXP <- "0"
data[!(data$CROPDMGEXP %in% fact_as_char), ]$CROPDMGEXP <- "0"Let's now have a look to the corrected results of multipliers. As you can see below, all wrong characters have now been replaced with character "0" and these 2 columns (PROPDMGEXP and CROPDMGEXP) have been cleaned meaning that they now only take permitted values, i.e. values included in the above-defined converting table.
table(data$PROPDMGEXP)##
## 0 1 2 3 4 5 6 7 8 B
## 466164 25 13 4 4 28 4 5 1 40
## H K M h m
## 6 424665 11330 1 7
table(data$CROPDMGEXP)##
## 0 2 B K M k m
## 618439 1 9 281832 1994 21 1
After having cleaned them in their string fomat, let's now convert the columns PROPDMGEXP and CROPDMGEXP into numerical multipliers according to the converting table defined above and let's calculate the properties and crop damages by multiplying the value of the damage (PROPDMG and CROPDMG) of each event with the corresponding numeric multiplier that we have just calculated for PROPDMGEXP and CROPDMGEXP columns. With this operation we have now calculated the total damage both to crop and to properties of each atmospheric event in the US from 1950 to 2011.
## Translate the string factors into numerical factors according to the
## conversion table. Caluclate total damages and save the results into 2
## news columns: data$PROPdamage and data$CROPdamage Notice that errors were
## replaced with character '0' which in turn are now converted in a numerical
## factor equal to 1.
data$PROPdamage <- data$PROPDMG * factors[match(data$PROPDMGEXP, factors$fact_as_char),
]$fact_as_num
data$CROPdamage <- data$CROPDMG * factors[match(data$CROPDMGEXP, factors$fact_as_char),
]$fact_as_num===========================================================================================
Let's now have a look to EVTYPE (event types) since what we want to do is summing all quantities of interest grouping by event type.
## Let's calculate the number of distinct EVTYPE Let's use only upper case to
## make this counting not case sensitive
data$EVTYPE <- toupper(data$EVTYPE)
events <- unique(data$EVTYPE)
unique_not_corrected_event_names <- unique(data$EVTYPE)
number_of_unique_not_corrected_event_names <- length(unique_not_corrected_event_names)In the intial data set 902297 atmospheric events are classified into 898 unique categories which is such a large number that would lead to results not easy to handle and to interpret. We need to define fewer but larger categories to group events and to improve the readability and interpretation of the results.
Additionally, a quick look to these names reveals that a categorization is needed also to clean initial categories from numerous errors and non-uniformities. For instance, in the initial dataset WIND, WND, WINT are 3 different, separated EVTYPEs. A non-uniform classification and typing errors make these 3 categories appear as they were different each other while they actually all represent "Wind" atmospheric events.
To solve the above-mentioned problems, a 2 steps categorization of EVTYPES will be now performed. The first automatic step will programmatically categorize EVTYPE depending on if it contains specific words. A second optional categorization will be later performed as a refinement of this first step. In other words, on optional basis, the second step will allow categorizing all those events, which were not categorized through the first step. You can skip this second step but if you do so, you will reach results similar to those of this study but not exactly the same since by skipping the refinement step, the resulting categorization would be more generic.
Let's start with the first categorization step:
- EVTYPES containing one of these words (WIND, WND, WINT) will be now classified as a Wind event type
- EVTYPES containing the word HAIL as Hail
- EVTYPES containing FLOOD as Flood
- EVTYPES containing TORNADO as Tornado
- EVTYPES containing LIGHTNING as Lightning
- EVTYPES containing (SNOW or ICE or COLD or BLIZZARD or ICY or FREEZ or FROST or ICE) as Low_Temperature
- EVTYPES containing (DRY or DROUGHT or WARMTH or HEAT or DRYNESS) as Hot_Dry
- EVTYPES containing (RAIN or PRECIPITATION or WET) as Rain
Word <- c("WIND", "WND", "WINT", "HAIL", "FLOOD", "TORNADO", "LIGHTNING", "SNOW",
"ICE", "COLD", "BLIZZARD", "ICY", "FREEZ", "FROST", "DRY", "DROUGHT", "WARMTH",
"HEAT", "DRYNESS", "RAIN", "PRECIPITATION")
Category <- c("Wind", "Wind", "Wind", "Hail", "Flood", "Tornado", "Lightning",
"Low_temperature", "Low_Temperature", "Low_Temperature", "Low_Temperature",
"Low_Temperature", "Low_Temperature", "Low_Temperature", "Hot_Dry", "Hot_Dry",
"Hot_Dry", "Hot_Dry", "Hot_Dry", "Rain", "Rain")
## Creates a conversion table for the first step categorization
first_classification <- data.frame(Word, Category)
first_classification$Word <- as.character(first_classification$Word)
first_classification$Category <- as.character(first_classification$Category)Let's display the classification table and let's use it to implement this first classification step.
first_classification## Word Category
## 1 WIND Wind
## 2 WND Wind
## 3 WINT Wind
## 4 HAIL Hail
## 5 FLOOD Flood
## 6 TORNADO Tornado
## 7 LIGHTNING Lightning
## 8 SNOW Low_temperature
## 9 ICE Low_Temperature
## 10 COLD Low_Temperature
## 11 BLIZZARD Low_Temperature
## 12 ICY Low_Temperature
## 13 FREEZ Low_Temperature
## 14 FROST Low_Temperature
## 15 DRY Hot_Dry
## 16 DROUGHT Hot_Dry
## 17 WARMTH Hot_Dry
## 18 HEAT Hot_Dry
## 19 DRYNESS Hot_Dry
## 20 RAIN Rain
## 21 PRECIPITATION Rain
## Let's create and initialize a new column in the daata frame, called
## EV_CATEGORY where we'll store the new classification.
data$EV_CATEGORY <- "***"
## Perform the first classification step: if EVTYPE contains a word of the
## first step conversion table, the corresponding category in the conversion
## table will be stored in the column EV_CATEGORY
for (i in 1:nrow(first_classification)) {
data[grepl(first_classification[i, ]$Word, data$EVTYPE, fixed = TRUE), ]$EV_CATEGORY <- first_classification[i,
]$Category
}
aggr <- aggregate(data[c("FATALITIES", "INJURIES", "PROPdamage", "CROPdamage")],
by = data[c("EV_CATEGORY")], FUN = sum)## Notice that those rows which still contains the initializing character ***
## in the EV_CATEGORY are the events which have not been classified yet.
## Let's calculate a few values about these events.
classified_fatalities <- sum(aggr[aggr$EV_CATEGORY != "***", ]$FATALITIES)
classified_injuries <- sum(aggr[aggr$EV_CATEGORY != "***", ]$INJURIES)
classified_CROPdamage <- sum(aggr[aggr$EV_CATEGORY != "***", ]$CROPdamage)
classified_PROPdamage <- sum(aggr[aggr$EV_CATEGORY != "***", ]$PROPdamage)
tot_fatalities <- sum(aggr$FATALITIES)
tot_injuries <- sum(aggr$INJURIES)
tot_CROPdamage <- sum(aggr$CROPdamage)
tot_PROPdamage <- sum(aggr$PROPdamage)
classified_number <- length(unique(data[data$EV_CATEGORY != "***", ]$EVTYPE))
uniq <- unique(data[data$EV_CATEGORY == "***", ]$EVTYPE)
unclassified_names_left <- length(uniq)
write.csv(uniq, "Unique unclassified events.csv")===========================================================================================
According to this first step classification, the initial 898 unique categories have been reduced since 621 of them were grouped into 21 more compelling categories. This categorization allowed to classify:
1)95.4 % of total injuries
2)89.7 % of total fatalities
3)86.1 % of total crop damages
4)64.7 % of total property damages into these new categories defined above.
To improve the amount of classified injuries, fatalities, properties and crop damages, a second, manual classification can be performed on optional basis on the remaining 277 events which have not been classified in the first step classification.
In the last line of the last code chunk, these unclassifies 277 events have been saved in a file ("Unique unclassified events.csv") for a subsequent optional manual classification to refine the final results.
The manual events classification applied to this study are listed in the Appendix 1.
At this point you have 2 options:
- If you want to reproduce exactly this study, you need to take the table in Appendix 1 and save those values in a csv file in your working directory. This file has to be saved as "Unique events after manual classification.csv", should not contain any header and should use ";" as values separator. By doing so, you will reach ecactly the same results as this study.
- Alternatively, you can skip this manual operation. The event types which were not classified in the first step will be all included in a generic category called Other. Results would be analogous to those of this study but less accurate.
In the next few lines, this second conversion table will be open and read in R and will be used to classify the 277 events which have not been classified in the first step classification.
## Let's now load the manually edited classification table which will allows
## us to classify the remaining events.
if (file.exists("Unique events after manual classification.csv")) {
## If the file exists, load the conversion table and apply the second step
## manual classification
Storm_classification <- read.table("Unique events after manual classification.csv",
sep = ";")
names(Storm_classification) <- c("Word", "Category")
Storm_classification$Category <- as.character(Storm_classification$Category)
Storm_classification$Word <- as.character(Storm_classification$Word)
for (i in 1:nrow(Storm_classification)) {
d <- data$EV_CATEGORY == "***" & data$EVTYPE == Storm_classification[i,
]$Word
data[d, ]$EV_CATEGORY <- Storm_classification[i, ]$Category
}
} else {
## If file does not exist, the second step manual classification will be
## skipped and the remaining events will be grouped in the Other category.
message("hol")
d <- data$EV_CATEGORY == "***"
data[d, ]$EV_CATEGORY <- "Other"
}
df <- aggregate(data[c("FATALITIES", "INJURIES", "PROPdamage", "CROPdamage")],
by = data[c("EV_CATEGORY")], FUN = sum)
classified_number <- length(unique(data[data$EV_CATEGORY != "***", ]$EVTYPE))The classification process of EVTYPE is now ended. All the initial 902297 events have been categorized into the following 18 categories.
df$EV_CATEGORY## [1] "Avalanche" "Duststorm" "Fire"
## [4] "Flood" "Hail" "Hot_Dry"
## [7] "Hurricane" "Landslide" "Lightning"
## [10] "Low_Temperature" "Low_temperature" "Other"
## [13] "Rain" "Sea" "Tornado"
## [16] "Typhoon" "Waterspout" "Wind"
===========================================================================================
As you can see, among the new categories, there is one called "Other" which contains all those events which could not be classified in other way.
Before proceeding with our analysis, let's evaluate the "weight" of the category "Other", compared with the rest of categories to make sure that it is small enough not to miss significant results for a too generic classification:
## Calculate the weight of 'Other' with respect of other categories in terms
## of fatalities, injuries and damages.
Other_fatalities <- sum(df[df$EV_CATEGORY == "Other", ]$FATALITIES)
Other_injuries <- sum(df[df$EV_CATEGORY == "Other", ]$INJURIES)
Other_CROPdamage <- sum(df[df$EV_CATEGORY == "Other", ]$CROPdamage)
Other_PROPdamage <- sum(df[df$EV_CATEGORY == "Other", ]$PROPdamage)
tot_fatalities <- sum(df$FATALITIES)
tot_injuries <- sum(df$INJURIES)
tot_CROPdamage <- sum(df$CROPdamage)
tot_PROPdamage <- sum(df$PROPdamage)
Non_other <- length(data[data$EV_CATEGORY != "Other", ]$EVTYPE)
Other <- length(data[data$EV_CATEGORY == "Other", ]$EVTYPE)Of the initial 902297 events, as low as 5640 events have been classified in the group Other while 896657 have been classified in more specific categories.
Additionally, the category "Other" represents:
- 0.98 % of total injuries
- 0.78 % of total fatalities
- 0.02 % of total crop damages
- 0.02 % of total proprerty damages
This numbers make me conclude that the category "Other"" is small enough and we do not need further classification refinements.
The number provided above are true under the assumption that you the manual classification of event type have not been skipped.
If you decided to skip the manual classification step of event types, all the events which have not been classified through the automated first classification step were included in the "Other" category. If you decided to skip the manual classification the numbers presented above would be therefore sensibily higher.
More specifically, in case of skipping the manual optional classification, the category "Other"" would represent:
- 4.62 % of total injuries
- 10.29 % of total fatalities
- 13.85 % of total crop damages
- 35.34 % of total proprerty damages
From now on, results will be shown only in the hypotesis that the second manual classification has been performed.
Let's now sum the variables of interest, grouping them by the new categories we have introduced. Finally let's plot the results. We will point out the 5 event types causing the highest number of fatalities, the top 5 categories causing the highest damages to crop and so on.
=========================================================================================== In this and following section, 3 combined plots will be creatd. Plase refer to the result section, to visualize them and read the comments on them.
## Let's first load a few library
library("ggplot2")
library("scales")
library("gridExtra")## Loading required package: grid
## Aggregate the variables of interest according to the calssification we
## made above.
df <- aggregate(data[c("FATALITIES", "INJURIES", "PROPdamage", "CROPdamage")],
by = data[c("EV_CATEGORY")], FUN = sum)
df$EV_CATEGORY <- as.character(df$EV_CATEGORY)
## Calculate the weight of each event with respect of all other events. An
## example to better explain: the tornado which hit the states in 1961, what
## percentage of injuries, fatalities etc have coused with respect of total
## injuries, fatalities etc since 1950?
df$Fat_percentage <- round(df$FATALITIES/sum(df$FATALITIES), 3)
df$Inj_percentage <- round(df$INJURIES/sum(df$INJURIES), 3)
df$CROP_percentage <- round(df$CROPdamage/sum(df$CROPdamage), 3)
df$PROP_percentage <- round(df$PROPdamage/sum(df$PROPdamage), 3)
## Inizializing the 4 new columns which will contain the percentage of each
## event with respect to the rest.
df$Fat_Cat <- "Others"
df$Inj_Cat <- "Others"
df$CROP_Cat <- "Others"
df$PROP_Cat <- "Others"
## Mark the top 5 event type per variable of interest
df <- df[order(-df$Fat_percentage), ]
df[1:5, ]$Fat_Cat <- df[1:5, ]$EV_CATEGORY
df <- df[order(-df$Inj_percentage), ]
df[1:5, ]$Inj_Cat <- df[1:5, ]$EV_CATEGORY
df <- df[order(-df$CROP_percentage), ]
df[1:5, ]$CROP_Cat <- df[1:5, ]$EV_CATEGORY
df <- df[order(-df$PROP_percentage), ]
df[1:5, ]$PROP_Cat <- df[1:5, ]$EV_CATEGORY
## Let's create the plots to display.
## Let's first disable de scientific notation for numbers
options(scipen = 999)
## Plot 1: fatalities
df1 <- aggregate(df[, c("Fat_percentage", "FATALITIES")], by = df[c("Fat_Cat")],
FUN = sum)
df1 <- df1[order(-df1$Fat_percentage), ]
n <- format(sum(df1$FATALITIES), big.mark = ",")
titl <- paste("Top 5 causes of FATALITIES\nTotal fatalities in the period = ",
n, sep = " ")
p1 <- ggplot(data = df1[df1$Fat_Cat != "Others", ], aes(x = reorder(Fat_Cat,
Fat_percentage), y = Fat_percentage)) + geom_bar(stat = "identity") + scale_size_area() +
xlab("") + ylab("As % of the total in the period") + ggtitle(titl) + scale_y_continuous(labels = percent) +
geom_text(aes(label = percent(Fat_percentage)), vjust = 1.5, colour = "white")
## Plot 2: injuries
df2 <- aggregate(df[, c("Inj_percentage", "INJURIES")], by = df[c("Inj_Cat")],
FUN = sum)
df2 <- df2[order(-df2$Inj_percentage), ]
n <- format(sum(df2$INJURIES), big.mark = ",")
titl <- paste("Top 5 causes of INJURIES\nTotal injuries in the period = ", n,
sep = " ")
p2 <- ggplot(data = df2[df2$Inj_Cat != "Others", ], aes(x = reorder(Inj_Cat,
Inj_percentage), y = Inj_percentage)) + geom_bar(stat = "identity") + scale_size_area() +
xlab("") + ylab("As % of the total in the period") + ggtitle(titl) + scale_y_continuous(labels = percent) +
geom_text(aes(label = percent(Inj_percentage)), vjust = 1.5, colour = "white")
## Plot 3: Property damages
df3 <- aggregate(df[, c("PROP_percentage", "PROPdamage")], by = df[c("PROP_Cat")],
FUN = sum)
df3 <- df3[order(-df3$PROP_percentage), ]
n <- format(round(sum(df3$PROPdamage), 0), big.mark = ",")
titl <- paste("Top 5 causes of PROPERTY DAMAGES\nTotal property damages in the period = ",
n, "$", sep = " ")
p3 <- ggplot(data = df3[df3$PROP_Cat != "Others", ], aes(x = reorder(PROP_Cat,
PROP_percentage), y = PROP_percentage)) + geom_bar(stat = "identity") +
scale_size_area() + xlab("") + ylab("As % of the total in the period") +
ggtitle(titl) + scale_y_continuous(labels = percent) + geom_text(aes(label = percent(PROP_percentage)),
vjust = 1.5, colour = "white")
## Plot 4: Crop damages
df4 <- aggregate(df$CROP_percentage, df$CROPdamage, by = df[c("CROP_Cat")],
FUN = sum)
df4 <- aggregate(df[, c("CROP_percentage", "CROPdamage")], by = df[c("CROP_Cat")],
FUN = sum)
df4 <- df4[order(-df4$CROP_percentage), ]
n <- format(round(sum(df4$CROPdamage), 0), big.mark = ",")
titl <- paste("Top 5 causes of CROP DAMAGES\nTotal crop damages in the period = ",
n, "$", sep = " ")
p4 <- ggplot(data = df4[df4$CROP_Cat != "Others", ], aes(x = reorder(CROP_Cat,
CROP_percentage), y = CROP_percentage)) + geom_bar(stat = "identity") +
scale_size_area() + xlab("") + ylab("As % of the total in the period") +
ggtitle(titl) + scale_y_continuous(labels = percent) + geom_text(aes(label = percent(CROP_percentage)),
vjust = 1.5, colour = "white")
## Create a title
mn = textGrob("MOST HARMFUL AND MOST DAMAGING WEATHER EVENTS IN THE US, 1950-2011\n",
gp = gpar(fontsize = 20))
## The plot has been built but it will be displayed in the results section===========================================================================================
## Let's now create the same graphs but split by decades.
## Defactorize the variable BGN_DATE and coerce it into a data format
data$BGN_DATE <- as.Date(data$BGN_DATE, format = "%m/%d/%Y")
## Calculate the 'decade' from the year stored into BGN_DATE
data$Decade <- as.character(trunc((as.numeric(format(data$BGN_DATE, "%Y")) -
1900)/10) * 10)
## A few adjustments
data[data$Decade == "100", ]$Decade <- "00"
data[data$Decade == "110", ]$Decade <- "10"
## Create a ordered factor with decades
data$Decade = factor(data$Decade, levels = c("50", "60", "70", "80", "90", "00",
"10"))
## Aggregating data of interest by decades
df <- aggregate(data[c("FATALITIES", "INJURIES", "PROPdamage", "CROPdamage")],
by = data[c("EV_CATEGORY", "Decade")], FUN = sum)
df2 <- aggregate(data[c("FATALITIES", "INJURIES", "PROPdamage", "CROPdamage")],
by = data[c("Decade")], FUN = sum)
## Create the plots
## Plot 1: fatalities
titl <- "FATALITIES in the US over decades"
p1b <- ggplot(data = df2, aes(x = Decade, y = FATALITIES)) + geom_bar(stat = "identity") +
scale_size_area() + xlab("") + ylab("Fatalities") + ggtitle(titl) + scale_y_continuous(labels = comma) +
geom_text(aes(label = format(round((FATALITIES), 0), big.mark = ",")), vjust = 1.5,
colour = "white")
## Plot 2: injuries
titl <- "INJURIES in the US over decades"
p2b <- ggplot(data = df2, aes(x = Decade, y = INJURIES)) + geom_bar(stat = "identity") +
scale_size_area() + xlab("") + ylab("Injuries") + ggtitle(titl) + scale_y_continuous(labels = comma) +
geom_text(aes(label = format(round((INJURIES), 0), big.mark = ",")), vjust = 1.5,
colour = "white")
# Plot 3: Property damages
titl <- "PROP DAMAGES in the US over decades in million US$"
p3b <- ggplot(data = df2, aes(x = Decade, y = PROPdamage)) + geom_bar(stat = "identity") +
scale_size_area() + xlab("") + ylab("Damages in million US$") + ggtitle(titl) +
scale_y_continuous(labels = comma) + geom_text(aes(label = format(round((PROPdamage/1000000),
0), big.mark = ",")), vjust = -0.2)
## Plot 4: Crop damages
titl <- "CROP DAMAGES in the US over decades in million US$"
p4b <- ggplot(data = df2, aes(x = Decade, y = CROPdamage)) + geom_bar(stat = "identity") +
scale_size_area() + xlab("") + ylab("Damages in US$ in million US$") + ggtitle(titl) +
scale_y_continuous(labels = comma) + geom_text(aes(label = format(round((CROPdamage/1000000),
0), big.mark = ",")), vjust = -0.2)
## Title of the plot
mnb = textGrob("DAMAGES, FATALITIES AND INJURIES EVOLUTION OVER DECADES IN THE US\n FROM THE 50s TO THE 10s (10s includes only the year 2011)",
gp = gpar(fontsize = 20))
===========================================================================================
## In this section, graphs referring to years 90s are generated. After the
## initial subsetting operation, the code is exactly the same as for the
## aggregated case already explained and commented above.
Nineties <- df[df$Decade == "90", ]
Nineties$EV_CATEGORY <- as.character(Nineties$EV_CATEGORY)
Nineties$Fat_percentage <- round(Nineties$FATALITIES/sum(Nineties$FATALITIES),
3)
Nineties$Inj_percentage <- round(Nineties$INJURIES/sum(Nineties$INJURIES), 3)
Nineties$CROP_percentage <- round(Nineties$CROPdamage/sum(Nineties$CROPdamage),
3)
Nineties$PROP_percentage <- round(Nineties$PROPdamage/sum(Nineties$PROPdamage),
3)
Nineties$Fat_Cat <- "Others"
Nineties$Inj_Cat <- "Others"
Nineties$CROP_Cat <- "Others"
Nineties$PROP_Cat <- "Others"
Nineties <- Nineties[order(-Nineties$Fat_percentage), ]
Nineties[1:5, ]$Fat_Cat <- Nineties[1:5, ]$EV_CATEGORY
Nineties <- Nineties[order(-Nineties$Inj_percentage), ]
Nineties[1:5, ]$Inj_Cat <- Nineties[1:5, ]$EV_CATEGORY
Nineties <- Nineties[order(-Nineties$CROP_percentage), ]
Nineties[1:5, ]$CROP_Cat <- Nineties[1:5, ]$EV_CATEGORY
Nineties <- Nineties[order(-Nineties$PROP_percentage), ]
Nineties[1:5, ]$PROP_Cat <- Nineties[1:5, ]$EV_CATEGORY
Nineties1 <- aggregate(Nineties[, c("Fat_percentage", "FATALITIES")], by = Nineties[c("Fat_Cat")],
FUN = sum)
Nineties1 <- Nineties1[order(-Nineties1$Fat_percentage), ]
n <- format(sum(Nineties1$FATALITIES), big.mark = ",")
titl <- paste("Top 5 causes of FATALITIES in the 90s\nTotal fatalities in the period = ",
n, sep = " ")
p_nineties_1 <- ggplot(data = Nineties1[Nineties1$Fat_Cat != "Others", ], aes(x = reorder(Fat_Cat,
Fat_percentage), y = Fat_percentage)) + geom_bar(stat = "identity") + scale_size_area() +
xlab("") + ylab("As % of the total in the period") + ggtitle(titl) + scale_y_continuous(labels = percent) +
geom_text(aes(label = percent(Fat_percentage)), vjust = 1.5, colour = "white")
Nineties2 <- aggregate(Nineties[, c("Inj_percentage", "INJURIES")], by = Nineties[c("Inj_Cat")],
FUN = sum)
Nineties2 <- Nineties2[order(-Nineties2$Inj_percentage), ]
n <- format(sum(Nineties2$INJURIES), big.mark = ",")
titl <- paste("Top 5 causes of INJURIES in the 90s\nTotal injuries in the period = ",
n, sep = " ")
p_nineties_2 <- ggplot(data = Nineties2[Nineties2$Inj_Cat != "Others", ], aes(x = reorder(Inj_Cat,
Inj_percentage), y = Inj_percentage)) + geom_bar(stat = "identity") + scale_size_area() +
xlab("") + ylab("As % of the total in the period") + ggtitle(titl) + scale_y_continuous(labels = percent) +
geom_text(aes(label = percent(Inj_percentage)), vjust = 1.5, colour = "white")
Nineties3 <- aggregate(Nineties[, c("PROP_percentage", "PROPdamage")], by = Nineties[c("PROP_Cat")],
FUN = sum)
Nineties3 <- Nineties3[order(-Nineties3$PROP_percentage), ]
n <- format(round(sum(Nineties3$PROPdamage), 0), big.mark = ",")
titl <- paste("Top 5 causes of PROPERTY DAMAGES in the 90s\nTotal property damages in the period = ",
n, "$", sep = " ")
p_nineties_3 <- ggplot(data = Nineties3[Nineties3$PROP_Cat != "Others", ], aes(x = reorder(PROP_Cat,
PROP_percentage), y = PROP_percentage)) + geom_bar(stat = "identity") +
scale_size_area() + xlab("") + ylab("As % of the total in the period") +
ggtitle(titl) + scale_y_continuous(labels = percent) + geom_text(aes(label = percent(PROP_percentage)),
vjust = 1.5, colour = "white")
Nineties4 <- aggregate(Nineties$CROP_percentage, Nineties$CROPdamage, by = Nineties[c("CROP_Cat")],
FUN = sum)
Nineties4 <- aggregate(Nineties[, c("CROP_percentage", "CROPdamage")], by = Nineties[c("CROP_Cat")],
FUN = sum)
Nineties4 <- Nineties4[order(-Nineties4$CROP_percentage), ]
n <- format(round(sum(Nineties4$CROPdamage), 0), big.mark = ",")
titl <- paste("Top 5 causes of CROP DAMAGES in the 90s\nTotal crop damages in the period = ",
n, "$", sep = " ")
p_nineties_4 <- ggplot(data = Nineties4[Nineties4$CROP_Cat != "Others", ], aes(x = reorder(CROP_Cat,
CROP_percentage), y = CROP_percentage)) + geom_bar(stat = "identity") +
scale_size_area() + xlab("") + ylab("As % of the total in the period") +
ggtitle(titl) + scale_y_continuous(labels = percent) + geom_text(aes(label = percent(CROP_percentage)),
vjust = 1.5, colour = "white")
===========================================================================================
## In this section, graphs for the years 2000s are created. After the
## initial subsetting operation, the code is exactly the same as for the
## aggregated case already explained and commented above.
Twothausands <- df[df$Decade == "00" | df$Decade == "10", ]
Twothausands <- aggregate(Twothausands[, c("FATALITIES", "INJURIES", "PROPdamage",
"CROPdamage")], by = Twothausands[c("EV_CATEGORY")], FUN = sum)
Twothausands$EV_CATEGORY <- as.character(Twothausands$EV_CATEGORY)
Twothausands$Fat_percentage <- round(Twothausands$FATALITIES/sum(Twothausands$FATALITIES),
3)
Twothausands$Inj_percentage <- round(Twothausands$INJURIES/sum(Twothausands$INJURIES),
3)
Twothausands$CROP_percentage <- round(Twothausands$CROPdamage/sum(Twothausands$CROPdamage),
3)
Twothausands$PROP_percentage <- round(Twothausands$PROPdamage/sum(Twothausands$PROPdamage),
3)
Twothausands$Fat_Cat <- "Others"
Twothausands$Inj_Cat <- "Others"
Twothausands$CROP_Cat <- "Others"
Twothausands$PROP_Cat <- "Others"
Twothausands <- Twothausands[order(-Twothausands$Fat_percentage), ]
Twothausands[1:5, ]$Fat_Cat <- Twothausands[1:5, ]$EV_CATEGORY
Twothausands <- Twothausands[order(-Twothausands$Inj_percentage), ]
Twothausands[1:5, ]$Inj_Cat <- Twothausands[1:5, ]$EV_CATEGORY
Twothausands <- Twothausands[order(-Twothausands$CROP_percentage), ]
Twothausands[1:5, ]$CROP_Cat <- Twothausands[1:5, ]$EV_CATEGORY
Twothausands <- Twothausands[order(-Twothausands$PROP_percentage), ]
Twothausands[1:5, ]$PROP_Cat <- Twothausands[1:5, ]$EV_CATEGORY
Twothausands1 <- aggregate(Twothausands[, c("Fat_percentage", "FATALITIES")],
by = Twothausands[c("Fat_Cat")], FUN = sum)
Twothausands1 <- Twothausands1[order(-Twothausands1$Fat_percentage), ]
n <- format(sum(Twothausands1$FATALITIES), big.mark = ",")
titl <- paste("Top 5 causes of FATALITIES in the 00s + 2011\nTotal fatalities in the period = ",
n, sep = " ")
p_Twothausands_1 <- ggplot(data = Twothausands1[Twothausands1$Fat_Cat != "Others",
], aes(x = reorder(Fat_Cat, Fat_percentage), y = Fat_percentage)) + geom_bar(stat = "identity") +
scale_size_area() + xlab("") + ylab("As % of the total in the period") +
ggtitle(titl) + scale_y_continuous(labels = percent) + geom_text(aes(label = percent(Fat_percentage)),
vjust = 1.5, colour = "white")
Twothausands2 <- aggregate(Twothausands[, c("Inj_percentage", "INJURIES")],
by = Twothausands[c("Inj_Cat")], FUN = sum)
Twothausands2 <- Twothausands2[order(-Twothausands2$Inj_percentage), ]
n <- format(sum(Twothausands2$INJURIES), big.mark = ",")
titl <- paste("Top 5 causes of INJURIES in the 00s + 2011\nTotal injuries in the period = ",
n, sep = " ")
p_Twothausands_2 <- ggplot(data = Twothausands2[Twothausands2$Inj_Cat != "Others",
], aes(x = reorder(Inj_Cat, Inj_percentage), y = Inj_percentage)) + geom_bar(stat = "identity") +
scale_size_area() + xlab("") + ylab("As % of the total in the period") +
ggtitle(titl) + scale_y_continuous(labels = percent) + geom_text(aes(label = percent(Inj_percentage)),
vjust = 1.5, colour = "white")
Twothausands3 <- aggregate(Twothausands[, c("PROP_percentage", "PROPdamage")],
by = Twothausands[c("PROP_Cat")], FUN = sum)
Twothausands3 <- Twothausands3[order(-Twothausands3$PROP_percentage), ]
n <- format(round(sum(Twothausands3$PROPdamage), 0), big.mark = ",")
titl <- paste("Top 5 causes of PROPERTY DAMAGES in the 00s + 2011\nTotal property damages in the period = ",
n, "$", sep = " ")
p_Twothausands_3 <- ggplot(data = Twothausands3[Twothausands3$PROP_Cat != "Others",
], aes(x = reorder(PROP_Cat, PROP_percentage), y = PROP_percentage)) + geom_bar(stat = "identity") +
scale_size_area() + xlab("") + ylab("As % of the total in the period") +
ggtitle(titl) + scale_y_continuous(labels = percent) + geom_text(aes(label = percent(PROP_percentage)),
vjust = 1.5, colour = "white")
Twothausands4 <- aggregate(Twothausands$CROP_percentage, Twothausands$CROPdamage,
by = Twothausands[c("CROP_Cat")], FUN = sum)
Twothausands4 <- aggregate(Twothausands[, c("CROP_percentage", "CROPdamage")],
by = Twothausands[c("CROP_Cat")], FUN = sum)
Twothausands4 <- Twothausands4[order(-Twothausands4$CROP_percentage), ]
n <- format(round(sum(Twothausands4$CROPdamage), 0), big.mark = ",")
titl <- paste("Top 5 causes of CROP DAMAGES in the 00s + 2011\nTotal crop damages in the period = ",
n, "$", sep = " ")
p_Twothausands_4 <- ggplot(data = Twothausands4[Twothausands4$CROP_Cat != "Others",
], aes(x = reorder(CROP_Cat, CROP_percentage), y = CROP_percentage)) + geom_bar(stat = "identity") +
scale_size_area() + xlab("") + ylab("As % of the total in the period") +
ggtitle(titl) + scale_y_continuous(labels = percent) + geom_text(aes(label = percent(CROP_percentage)),
vjust = 1.5, colour = "white")
mnc = textGrob("MOST HARMFUL AND MOST DAMAGING WEATHER EVENTS IN THE US\nComparison between the 90s vs 00s+2011",
gp = gpar(fontsize = 20))
===========================================================================================
Let's see which were the worst 5 types of waether events for each variable of interest. We will presents the result by commenting the 3 following combined graphs:
- 1 combinined graph showing the top 5 causes of damage to property, damage to crop, injuries and fatalities in the 1950-2011 period
- 1 combinined graph showing the evolution across decades of damage to crop and to property, injuries and fatalities
- 1 combinined graph comparing the 5 causes of damage to property, damage to crop, injuries and fatalities in the 90s versus the corresponding causes in the 2000s
## Display 1950-2011 plot
grid.arrange(p1, p2, p3, p4, nrow = 2, main = mn)
If we consider the whole period from the 1950 to 2011, Tornado has been the most harmful weather event in the US causing on its own nearly 40% of total fatalities and up to 65% of total injuries in the period for weather conditions.
However the graph also shows that Tornado is not the most destructive atmospheric event: it is "only" the third major cause of property damage and it is not even in the top 5 weather causes of highest damage to crop.
Hot_dry was the type of weather event causing of highest damage to crop accounting for nearly a third of all dmage to crops since 1950.
Flood is even worse since not only it is the event causing the highest damage to properties (nearly 40%) but it is also the event causing the second highest damage to crops accounting for a 25% of all damages..
## Display the by decades plot
grid.arrange(p1b, p2b, p3b, p4b, nrow = 2, main = mnb)
From the analysis of data split by decades, it is clear that the 90s and in the 2000s is the time period where damages, fatalities and injuries are concentrated the most.
Of course this does not mean that in the 60s, there were less dangerous and less destructive events events. I guess that the reason behind this strange concentration on the most recent decades is because data started to be systematically and more accurately gathered starting from the 90s. This is probably why, till the 90s crop damages were 0 and strated increasing since the 90s. This should rather be interpreted as missing values rather than 0 damage.
For what it has been said above, it may be interesting comparing data from the 90s with data from the 2000s (including 2011 which is included in the 2000s for completeness: 1 year alone is not enough to consider a new decade but on the other hand, it is not even correct ignoring the year 2011)
## Display 90vs 00s plot
grid.arrange(p_nineties_1, p_Twothausands_1, p_nineties_2, p_Twothausands_2,
p_nineties_3, p_Twothausands_3, p_nineties_4, p_Twothausands_4, ncol = 2,
main = mnc)
Observing data from the 90s and 2000s, it is immediately evident that Tornado was not the most harmful event for human beings. Hot_dry caused more fatalities than Tornado (38% of fatalities in the 90s and 21% in the 2000s have been caused by hot dry while tornados caused the 11,5% and 20% of fatalities in the 90s and 2000s respectively). Hot dry events were also the main enemy of crops in the 2000s while it ranks as 3rd major cause of damages to crop in the 90s behind low temperature and flood.
For what it concerns properties damages in the 90s and 2000s flood confirms to be the weather event causing highest amount of damage, much higher than the second worse atmospheric event type (Flood in the 90s caused 30% of total damages to properties while Tornado, the second most destructive event accounted for 17.8%; in yje 2000s, Flood accounted for 44% of total damages while the second worse event, hurricanes, accounted for less than a half.
===========================================================================================
Storm_classification## Word Category
## 1 HURRICANE ERIN Hurricane
## 2 HURRICANE OPAL Hurricane
## 3 DENSE FOG Other
## 4 RIP CURRENT Sea
## 5 THUNDERSTORM WINS Lightning
## 6 FUNNEL CLOUD Rain
## 7 LIGHTING Lightning
## 8 FUNNEL Other
## 9 WALL CLOUD Rain
## 10 WALL CLOUD/FUNNEL CLOUD Rain
## 11 THUNDERSTORM Lightning
## 12 WATERSPOUT Waterspout
## 13 HIGH TIDES Other
## 14 RECORD HIGH TEMPERATURE Hot_Dry
## 15 RECORD HIGH Hot_Dry
## 16 RECORD LOW Low_Temperature
## 17 LOW TEMPERATURE RECORD Low_Temperature
## 18 AVALANCHE Avalanche
## 19 MARINE MISHAP Sea
## 20 HIGH TEMPERATURE RECORD Hot_Dry
## 21 RECORD HIGH TEMPERATURES Hot_Dry
## 22 HIGH SEAS Sea
## 23 SEVERE TURBULENCE Other
## 24 DUST STORM Duststorm
## 25 APACHE COUNTY Other
## 26 SLEET Low_Temperature
## 27 FUNNEL CLOUDS Rain
## 28 DUST DEVIL Duststorm
## 29 HEAVY SURF Sea
## 30 HEAVY PRECIPATATION Rain
## 31 HIGH SURF Sea
## 32 BLOWING DUST Duststorm
## 33 URBAN/SMALL Other
## 34 WILD FIRES Fire
## 35 HIGH Other
## 36 WATER SPOUT Waterspout
## 37 MUDSLIDES Other
## 38 SEVERE THUNDERSTORM Lightning
## 39 SEVERE THUNDERSTORMS Lightning
## 40 THUNDERSTORMS Lightning
## 41 MICROBURST Other
## 42 URBAN AND SMALL Other
## 43 DOWNBURST Other
## 44 GUSTNADO AND Tornado
## 45 WET MICROBURST Rain
## 46 GLAZE Other
## 47 UNSEASONABLY WET Rain
## 48 RIP CURRENTS HEAVY SURF Sea
## 49 UNSEASONABLY WARM Hot_Dry
## 50 WATERSPOUTS Waterspout
## 51 STORM SURGE Sea
## 52 WATERSPOUT- Waterspout
## 53 TROPICAL STORM ALBERTO Rain
## 54 TROPICAL STORM Rain
## 55 TROPICAL STORM GORDON Rain
## 56 TROPICAL STORM JERRY Rain
## 57 WAYTERSPOUT Sea
## 58 URBAN AND SMALL STREAM Other
## 59 WILDFIRE Fire
## 60 HURRICANE Hurricane
## 61 WILD/FOREST FIRE Fire
## 62 MUD SLIDE Landslide
## 63 LIGNTNING Lightning
## 64 COOL AND WET Rain
## 65 SMALL STREAM AND Other
## 66 MUD SLIDES Other
## 67 EXCESSIVE WETNESS Rain
## 68 ROTATING WALL CLOUD Rain
## 69 LARGE WALL CLOUD Rain
## 70 GUSTNADO Tornado
## 71 FOG Other
## 72 THUNDERSTORM DAMAGE Lightning
## 73 GRASS FIRES Fire
## 74 THUNDERSTORMW 50 Other
## 75 TROPICAL STORM DEAN Rain
## 76 HURRICANE-GENERATED SWELLS Hurricane
## 77 COASTAL SURGE Sea
## 78 HIGH WAVES Sea
## 79 DUST DEVIL WATERSPOUT Waterspout
## 80 WATERSPOUT/ Waterspout
## 81 URBAN/SMALL STREAM Other
## 82 FUNNEL CLOUD. Rain
## 83 TORNDAO Tornado
## 84 AVALANCE Avalanche
## 85 FOREST FIRES Fire
## 86 WILDFIRES Fire
## 87 RIP CURRENTS/HEAVY SURF Sea
## 88 RECORD TEMPERATURES Hot_Dry
## 89 OTHER Other
## 90 MUDSLIDE Landslide
## 91 HEAVY MIX Other
## 92 RIP CURRENTS Sea
## 93 HURRICANE EMILY Hurricane
## 94 HURRICANE GORDON Hurricane
## 95 HURRICANE FELIX Hurricane
## 96 THUNDERSTORM DAMAGE TO Lightning
## 97 DAM FAILURE Other
## 98 SOUTHEAST Other
## 99 HIGH WATER Sea
## 100 WET WEATHER Rain
## 101 BEACH EROSIN Sea
## 102 LOW TEMPERATURE Low_Temperature
## 103 THUNDERSTORM W INDS Lightning
## 104 HYPOTHERMIA Low_Temperature
## 105 MUD/ROCK SLIDE Landslide
## 106 RAPIDLY RISING WATER Other
## 107 FLASH FLOOODING Flood
## 108 THUNDERSTORMW Lightning
## 109 TSTMW Other
## 110 FUNNELS Other
## 111 LANDSLIDE Landslide
## 112 EXCESSIVE Other
## 113 WILD/FOREST FIRES Fire
## 114 HEAVY SEAS Sea
## 115 DUSTSTORM Duststorm
## 116 ? Other
## 117 HOT PATTERN Hot_Dry
## 118 BRUSH FIRES Fire
## 119 MILD PATTERN Other
## 120 LANDSLIDES Landslide
## 121 HEAVY SHOWERS Rain
## 122 BRUSH FIRE Fire
## 123 WATERSPOUT FUNNEL CLOUD Waterspout
## 124 SAHARAN DUST Duststorm
## 125 HEAVY SHOWER Rain
## 126 HEAVY SWELLS Sea
## 127 URBAN SMALL Other
## 128 SMALL STREAM Other
## 129 URBAN/SML STREAM FLD Other
## 130 TEMPERATURE RECORD Hot_Dry
## 131 ROUGH SURF Sea
## 132 MARINE ACCIDENT Other
## 133 COASTAL STORM Sea
## 134 WET MONTH Rain
## 135 WET YEAR Rain
## 136 BEACH EROSION Sea
## 137 LANDSLUMP Landslide
## 138 HURRICANE EDOUARD Hurricane
## 139 RECORD WARM TEMPS. Hot_Dry
## 140 RECORD TEMPERATURE Hot_Dry
## 141 MIXED PRECIP Rain
## 142 SUMMARY JAN 17 Other
## 143 SUMMARY OF MARCH 14 Other
## 144 SUMMARY OF MARCH 23 Other
## 145 SUMMARY OF MARCH 24 Other
## 146 SUMMARY OF APRIL 3RD Other
## 147 SUMMARY OF APRIL 12 Other
## 148 SUMMARY OF APRIL 13 Other
## 149 SUMMARY OF APRIL 21 Other
## 150 SUMMARY AUGUST 11 Other
## 151 SUMMARY OF APRIL 27 Other
## 152 SUMMARY OF MAY 9-10 Other
## 153 SUMMARY OF MAY 10 Other
## 154 SUMMARY OF MAY 13 Other
## 155 SUMMARY OF MAY 14 Other
## 156 SUMMARY OF MAY 22 AM Other
## 157 SUMMARY OF MAY 22 PM Other
## 158 SUMMARY OF MAY 26 AM Other
## 159 SUMMARY OF MAY 26 PM Other
## 160 METRO STORM, MAY 26 Other
## 161 SUMMARY OF MAY 31 AM Other
## 162 SUMMARY OF MAY 31 PM Other
## 163 SUMMARY OF JUNE 3 Other
## 164 SUMMARY OF JUNE 4 Other
## 165 SUMMARY JUNE 5-6 Other
## 166 SUMMARY JUNE 6 Other
## 167 SUMMARY OF JUNE 11 Other
## 168 SUMMARY OF JUNE 12 Other
## 169 SUMMARY OF JUNE 13 Other
## 170 SUMMARY OF JUNE 15 Other
## 171 SUMMARY OF JUNE 16 Other
## 172 SUMMARY JUNE 18-19 Other
## 173 SUMMARY OF JUNE 23 Other
## 174 SUMMARY OF JUNE 24 Other
## 175 SUMMARY OF JUNE 30 Other
## 176 SUMMARY OF JULY 2 Other
## 177 SUMMARY OF JULY 3 Other
## 178 SUMMARY OF JULY 11 Other
## 179 SUMMARY OF JULY 22 Other
## 180 SUMMARY JULY 23-24 Other
## 181 SUMMARY OF JULY 26 Other
## 182 SUMMARY OF JULY 29 Other
## 183 SUMMARY OF AUGUST 1 Other
## 184 SUMMARY AUGUST 2-3 Other
## 185 SUMMARY AUGUST 7 Other
## 186 SUMMARY AUGUST 9 Other
## 187 SUMMARY AUGUST 10 Other
## 188 SUMMARY AUGUST 17 Other
## 189 SUMMARY AUGUST 21 Other
## 190 SUMMARY AUGUST 28 Other
## 191 SUMMARY SEPTEMBER 4 Other
## 192 SUMMARY SEPTEMBER 20 Other
## 193 SUMMARY SEPTEMBER 23 Other
## 194 SUMMARY SEPT. 25-26 Other
## 195 SUMMARY: OCT. 20-21 Other
## 196 SUMMARY: OCTOBER 31 Other
## 197 SUMMARY: NOV. 6-7 Other
## 198 SUMMARY: NOV. 16 Other
## 199 WET MICOBURST Other
## 200 NO SEVERE WEATHER Other
## 201 SUMMARY OF MAY 22 Other
## 202 SUMMARY OF JUNE 6 Other
## 203 SUMMARY AUGUST 4 Other
## 204 SUMMARY OF JUNE 10 Other
## 205 SUMMARY OF JUNE 18 Other
## 206 SUMMARY SEPTEMBER 3 Other
## 207 SUMMARY: SEPT. 18 Other
## 208 SML STREAM FLD Other
## 209 MUDSLIDE/LANDSLIDE Other
## 210 VOLCANIC ASH Other
## 211 VOLCANIC ASH PLUME Other
## 212 NONE Other
## 213 DAM BREAK Other
## 214 BLOW-OUT TIDES Sea
## 215 UNSEASONABLY COOL Low_Temperature
## 216 URBAN/SML STREAM FLDG Other
## 217 BLOW-OUT TIDE Sea
## 218 HYPOTHERMIA/EXPOSURE Low_Temperature
## 219 COASTALSTORM Sea
## 220 SUMMARY OF MARCH 24-25 Other
## 221 SUMMARY OF MARCH 27 Other
## 222 SUMMARY OF MARCH 29 Other
## 223 URBAN/SMALL STRM FLDG Other
## 224 TYPHOON Typhoon
## 225 HIGH SWELLS Sea
## 226 HIGH SWELLS Sea
## 227 HOT SPELL Hot_Dry
## 228 UNSEASONABLY HOT Hot_Dry
## 229 COASTAL EROSION Sea
## 230 SEICHE Sea
## 231 TSTM Other
## 232 UNSEASONABLY WARM YEAR Hot_Dry
## 233 HYPERTHERMIA/EXPOSURE Hot_Dry
## 234 ROCK SLIDE Other
## 235 PATCHY DENSE FOG Other
## 236 RECORD COOL Low_Temperature
## 237 RECORD WARM Hot_Dry
## 238 HOT WEATHER Hot_Dry
## 239 TROPICAL DEPRESSION Other
## 240 VOLCANIC ERUPTION Other
## 241 COOL SPELL Low_Temperature
## 242 RED FLAG FIRE WX Fire
## 243 VOG Other
## 244 MONTHLY TEMPERATURE Other
## 245 REMNANTS OF FLOYD Other
## 246 LANDSPOUT Tornado
## 247 DRIEST MONTH Hot_Dry
## 248 ROUGH SEAS Sea
## 249 UNSEASONABLY WARM/WET Rain
## 250 UNSEASONABLY COOL & WET Rain
## 251 UNUSUALLY WARM Hot_Dry
## 252 WARM WEATHER Hot_Dry
## 253 UNSEASONABLY WARM & WET Rain
## 254 UNSEASONAL LOW TEMP Low_Temperature
## 255 HIGH SURF ADVISORY Sea
## 256 RED FLAG CRITERIA Other
## 257 SMOKE Other
## 258 WATERSPOUT Waterspout
## 259 EXTREMELY WET Rain
## 260 ROGUE WAVE Sea
## 261 DUST DEVEL Duststorm
## 262 NORTHERN LIGHTS Rain
## 263 HIGH SURF ADVISORY Sea
## 264 HAZARDOUS SURF Sea
## 265 ASTRONOMICAL HIGH TIDE Sea
## 266 VERY WARM Hot_Dry
## 267 ABNORMALLY WET Rain
## 268 DROWNING Hot_Dry
## 269 HIGH SURF ADVISORIES Sea
## 270 HURRICANE/TYPHOON Hurricane
## 271 HEAVY SURF/HIGH SURF Sea
## 272 SLEET STORM Low_Temperature
## 273 STORM SURGE/TIDE Sea
## 274 TSUNAMI Sea
## 275 DENSE SMOKE Other
## 276 ASTRONOMICAL LOW TIDE Sea
## 277 VOLCANIC ASHFALL Other
===========================================================================================