Hurricane Analysis

1. Introduction

Let’s perform an analysis on tropical cyclones in the North Atlantic. The data that we’ll be using is from the International Best Track Archive for Climate Stewardship. The CSV file can be found here while the data documentation can be found here.

Tropical cyclones are also known as tropical systems. They generally form near the tropical regions where the ocean is at least 80 degrees Fahrenheit. This warm water evaporates which creates warm moist air that acts as fuel for the storms. When this warm air rises into the atmosphere it begins to cool and large stormy clouds begin to form. As more warm air continues to rise then the winds will spiral and increase in speed.

The wind speed in the tropical cyclones is used to classify each storm. According to the Saffir-Simpson Wind Scale we have the following table

Tropical Cyclone	Sustained Wind Speed	Type of Hurricane	Named
Tropical Depression	38 mph or less	Not a hurricane	Unnamed
Tropical Storm	39 - 73 mph	Not a hurricane	Named
Category 1	74 - 95 mph	Hurricane	Named
Category 2	96 - 110 mph	Hurricane	Named
Category 3	111 - 129 mph	Major hurricane	Named
Category 4	130 - 156 mph	Major hurricane	Named
Category 5	157 mph or higher	Major hurricane	Named

It’s also important to note that hurricanes, cyclones, and typhoons are all tropical cyclones but are referred to as different names depending on the location in which they form. Hurricanes form in the North Atlantic, central North Pacific, and eastern North Pacific. Cyclones form in the South Pacific and Indian Ocean. Finally, typhoons form in the Northwest Pacific. More information on hurricanes form can be found at the University Corporation For Atmospheric Research.

2. Data Preparation

To begin working with our dataset we’re going to need the following R packages

Code

library(tidyverse)
library(lubridate)
library(spData)
library(rmdformats)

Code

original_DF = read.csv(file = "./data/ibtracs.NA.list.v04r00.csv")

Code

head(select(original_DF,1:6))

            SID SEASON NUMBER BASIN SUBBASIN      NAME
1                 Year     NA                         
2 1851175N26270   1851      5  <NA>       GM NOT_NAMED
3 1851175N26270   1851      5  <NA>       GM NOT_NAMED
4 1851175N26270   1851      5  <NA>       GM NOT_NAMED
5 1851175N26270   1851      5  <NA>       GM NOT_NAMED
6 1851175N26270   1851      5  <NA>       GM NOT_NAMED

We can see that the first row is unnecessary. We can also see that in the column BASIN the entry NA is being returned as a missing value in R, but it actually represents the North Atlantic region. Therefore we should fix this.

Now we can clean our data by changing the column names, specifying the data types, and by specifying how missing values are encoded

Code

# Vector of names for first 16 columns

col_names = c("SID",
              "SEASON",
              "NUMBER",
              "BASIN",
              "SUBBASIN",
              "NAME",
              "ISO_TIME",
              "NATURE",
              "LAT",
              "LON",
              "WMO_WIND",
              "WMO_PRES",
              "WMO_AGENCY",
              "TRACK_TYPE",
              "DIST2LAND",
              "LANDFALL"
              )

# Vector of data-types for first 16 columns

col_types = c("character",
              "integer",
              "integer",
              "character",
              "character",
              "character",
              "character",
              "character",
              "numeric",
              "numeric",
              "integer",
              "integer",
              "character",
              "character",
              "integer",
              "integer"
              )

DF = read.csv(file = "./data/ibtracs.NA.list.v04r00.csv",
                stringsAsFactors = FALSE,
                skip = 1,
                na.strings = ""
                )

# Renaming columns using col_names

colnames(DF) = col_names

# Update df to only the first 16 columns

DF = select(DF, 1:16)

In our analysis it will also be useful work with months so let’s add a month column to our dataset

Code

DF$ISO_TIME = as.POSIXct(DF$ISO_TIME)
DF$MONTH = lubridate::month(DF$ISO_TIME)

We’ll also focus on the time period beginning from 1970 instead of 1851 because in the 1960’s we began to track these storms with satellites.

Code

DF = filter(DF, SEASON >= 1970)

Now let’s take a look at our data

Code

head(select(DF,1:6))

            SID SEASON NUMBER BASIN SUBBASIN NAME
1 1970138N12281   1970     43    NA       CS ALMA
2 1970138N12281   1970     43    NA       CS ALMA
3 1970138N12281   1970     43    NA       CS ALMA
4 1970138N12281   1970     43    NA       CS ALMA
5 1970138N12281   1970     43    NA       CS ALMA
6 1970138N12281   1970     43    NA       CS ALMA

Code

str(DF, vec.len = 1)

'data.frame':   47100 obs. of  17 variables:
 $ SID       : chr  "1970138N12281" ...
 $ SEASON    : int  1970 1970 ...
 $ NUMBER    : int  43 43 ...
 $ BASIN     : chr  "NA" ...
 $ SUBBASIN  : chr  "CS" ...
 $ NAME      : chr  "ALMA" ...
 $ ISO_TIME  : POSIXct, format: "1970-05-17 18:00:00" ...
 $ NATURE    : chr  "TS" ...
 $ LAT       : num  11.5 ...
 $ LON       : num  -79 ...
 $ WMO_WIND  : int  25 NA ...
 $ WMO_PRES  : int  NA NA ...
 $ WMO_AGENCY: chr  "hurdat_atl" ...
 $ TRACK_TYPE: chr  "main" ...
 $ DIST2LAND : int  224 235 ...
 $ LANDFALL  : int  224 235 ...
 $ MONTH     : num  5 5 ...

Now we can begin working with our dataset.

3. Analysis

We’ll be focusing on the 2012 Atlantic hurricane season. Let’s create a new that includes all of the depressions that occurred in 2012. Therefore we have the following

Code

depressions_2012 = filter(DF, SEASON == 2012)

3.1. Seasonal Statistics

3.1.a.) Total number of depressions

A tropical depression is such that are those that have winds of 38 mph or less. Therefore we can find the total with the following code

Code

# 38 mph = 33.02 knots

depressions_2012 %>%
  filter(WMO_WIND <= 33.02) %>%
  group_by(NAME) %>%
  summarise(SID = first(SID),
            BASIN = first(BASIN))

# A tibble: 19 × 3
   NAME     SID           BASIN
   <chr>    <chr>         <chr>
 1 ALBERTO  2012140N33283 NA   
 2 BERYL    2012147N30284 NA   
 3 CHRIS    2012169N29291 NA   
 4 DEBBY    2012176N26272 NA   
 5 ERNESTO  2012215N12313 NA   
 6 FLORENCE 2012216N12337 NA   
 7 GORDON   2012229N28305 NA   
 8 HELENE   2012223N14317 NA   
 9 ISAAC    2012234N16315 NA   
10 JOYCE    2012235N11328 NA   
11 KIRK     2012242N24317 NA   
12 LESLIE   2012242N13333 NA   
13 MICHAEL  2012246N29323 NA   
14 NADINE   2012255N16322 NA   
15 OSCAR    2012277N12322 NA   
16 PATTY    2012285N26288 NA   
17 RAFAEL   2012287N15297 NA   
18 SANDY    2012296N14283 NA   
19 TONY     2012296N20309 NA   

Therefore we have that the total number of depressions was 19 in 2012.

3.1.b.) Number of named storms

We know that the number of named storms are those such that have winds of at least 39 mph. Therefore we can find the total with the following code

Code

# 39 mph = 33.89 knots

depressions_2012 %>%
  group_by(NAME) %>%
  filter(WMO_WIND > 33.89) %>%
  summarise(SID = first(SID),
            BASIN = first(BASIN))

# A tibble: 19 × 3
   NAME     SID           BASIN
   <chr>    <chr>         <chr>
 1 ALBERTO  2012140N33283 NA   
 2 BERYL    2012147N30284 NA   
 3 CHRIS    2012169N29291 NA   
 4 DEBBY    2012176N26272 NA   
 5 ERNESTO  2012215N12313 NA   
 6 FLORENCE 2012216N12337 NA   
 7 GORDON   2012229N28305 NA   
 8 HELENE   2012223N14317 NA   
 9 ISAAC    2012234N16315 NA   
10 JOYCE    2012235N11328 NA   
11 KIRK     2012242N24317 NA   
12 LESLIE   2012242N13333 NA   
13 MICHAEL  2012246N29323 NA   
14 NADINE   2012255N16322 NA   
15 OSCAR    2012277N12322 NA   
16 PATTY    2012285N26288 NA   
17 RAFAEL   2012287N15297 NA   
18 SANDY    2012296N14283 NA   
19 TONY     2012296N20309 NA   

Therefore we have that the total number of named storms was 19 in 2012.

3.1.c.) The number of hurricanes, and their names

We know that a named storm becomes a hurricane once its winds reach at least 74 mph. Therefore we can find the total with the following code

Code

# 74 mph = 64.30 knots

depressions_2012 %>%
  filter(WMO_WIND >= 64.30) %>%
  group_by(NAME) %>%
  summarise(SID = first(SID),
            BASIN = first(BASIN))

# A tibble: 10 × 3
   NAME    SID           BASIN
   <chr>   <chr>         <chr>
 1 CHRIS   2012169N29291 NA   
 2 ERNESTO 2012215N12313 NA   
 3 GORDON  2012229N28305 NA   
 4 ISAAC   2012234N16315 NA   
 5 KIRK    2012242N24317 NA   
 6 LESLIE  2012242N13333 NA   
 7 MICHAEL 2012246N29323 NA   
 8 NADINE  2012255N16322 NA   
 9 RAFAEL  2012287N15297 NA   
10 SANDY   2012296N14283 NA   

Therefore we have that the number of hurricanes was 10 in 2012.

3.1.d.) Number of major hurricanes, and their names

We know that a storm becomes a major hurricane when its wind speeds reach 111 mph. We can find this total with the following code

Code

# 111 mph = 96.46 knots

depressions_2012 %>%
  filter(WMO_WIND >= 96.46) %>%
  group_by(NAME) %>% 
  summarise(SID = first(SID),
            BASIN = first(BASIN))

# A tibble: 2 × 3
  NAME    SID           BASIN
  <chr>   <chr>         <chr>
1 MICHAEL 2012246N29323 NA   
2 SANDY   2012296N14283 NA   

Therefore we can conclude that there were two major hurricanes in 2012 and there names were Michael and Sandy.

3.1.e.) Date of the first system formed, and its name.

We can find this with the following code

Code

depressions_2012 %>%
  select(NAME, ISO_TIME) %>%
  arrange(ISO_TIME) %>%
  slice(1)

     NAME   ISO_TIME
1 ALBERTO 2012-05-19

Therefore the date of the first system was 2012-05-19 and its name was Alberto.

3.1.f.) Date of the last system dissipated, and its name

We can find this with the following code

Code

depressions_2012 %>%
  select(NAME, ISO_TIME) %>%
  arrange(desc(ISO_TIME)) %>%
  slice(1)

   NAME            ISO_TIME
1 SANDY 2012-10-31 12:00:00

Therefore the date of the last system dissipated was 2012-10-31 and its name was Sandy.

3.2. Statistics of Strongest Storm

3.2.a.) Name of the strongest storm

We can find this with the following

Code

depressions_2012 %>%
  filter(WMO_WIND == max(depressions_2012$WMO_WIND, na.rm = TRUE)) %>%
  group_by(NAME) %>%
  summarise()

# A tibble: 2 × 1
  NAME   
  <chr>  
1 MICHAEL
2 SANDY  

Therefore we can conclude that there were two storms that were equally as strong. These storms were Michael and Sandy.

3.2.b.) Its maximum wind speed

We can find this with the following

Code

depressions_2012 %>%
  filter(WMO_WIND == max(WMO_WIND, na.rm = TRUE)) %>%
  group_by(NAME) %>%
  summarise(Max_Wind =  max(WMO_WIND, na.rm = TRUE))

# A tibble: 2 × 2
  NAME    Max_Wind
  <chr>      <int>
1 MICHAEL      100
2 SANDY        100

The maximum wind speed for both Michael and Sandy were 100 knots or about 115 mph.

3.2.c.) Its lowest pressure

We can find this with the following code

Code

depressions_2012 %>%
  filter(NAME == "MICHAEL" | NAME == "SANDY") %>%
  group_by(NAME) %>%
  summarise(min_pres = min(WMO_PRES, na.rm = TRUE),
            max_pres = max(WMO_PRES, na.rm = TRUE)
            )

# A tibble: 2 × 3
  NAME    min_pres max_pres
  <chr>      <int>    <int>
1 MICHAEL      964     1015
2 SANDY        940     1006

Therefore the lowest pressure for Michael was 964, and the lowest pressure for Sandy was 940.

3.3. Historical Comparison

3.3.a.) Compared to the “historical” average number of named storms, was the 2012 season an above-average, a below-average, or an average season?

First, let’s create a data frame named_storms which contains all of the named storms between 1970 and 2021. Therefore we have

Code

# 39 mph = 33.89 knots

named_storms = DF %>%
  group_by(NAME) %>%
  select(NAME, SEASON, WMO_WIND) %>%
  filter(WMO_WIND >= 33.89) %>%
  count(SEASON)

We can display the first 5 rows of this data frame

Code

head(named_storms)

# A tibble: 6 × 3
# Groups:   NAME [2]
  NAME    SEASON     n
  <chr>    <int> <int>
1 AGNES     1972    21
2 ALBERTO   1982     7
3 ALBERTO   1988     4
4 ALBERTO   1994     9
5 ALBERTO   2000    82
6 ALBERTO   2006    28

Now using named_storms we can create a new data frame num_mamed_storm which contains all of the named storms that happened per year

Code

num_named_storms = named_storms %>%
  group_by(SEASON) %>%
  count(SEASON)

We can display the first 5 rows of num_named_storms

Code

head(num_named_storms)

# A tibble: 6 × 2
# Groups:   SEASON [6]
  SEASON     n
   <int> <int>
1   1970     8
2   1971    13
3   1972     7
4   1973     8
5   1974     8
6   1975     9

Now we can compute the average number of named storms

Code

mean(num_named_storms$n)

[1] 12.55769

We can extract the total number of named storms in 2012 with the following

Code

filter(num_named_storms, SEASON == 2012)

# A tibble: 1 × 2
# Groups:   SEASON [1]
  SEASON     n
   <int> <int>
1   2012    19

Therefore the average number of named storms between 1970 and 2021 was about 12 per year. In 2012 there were 19 named storms, therefore 2012 was an above-average season for named storms.

We can see this in the following graph

Code

ggplot(data = num_named_storms, aes(x = SEASON, y = n)) +
    geom_col(width = .8,
             fill = my_grey1) + 
    labs(title = "Total Number of Named Storms",
         subtitle = "Comparing the historical average number of named storms to the 2012 season",
         x = "Year",
         y = "Total") + 
  geom_col(data = filter(num_named_storms, SEASON == 2012),
           width = .8,
           fill = my_red) + 
  geom_text(x = 2012,
            y = 21,
            label = "2012",
            color = my_red,
            size = 6.8) + 
  geom_hline(yintercept = mean(num_named_storms$n),
             color = my_blue,
             size = 1) + 
  geom_text(x = 1974,
            y = 20,
            size = 6,
            color = my_blue,
            aes(label = "Average")) + 
  geom_segment(x = 1974, y = 18.3,
               xend = 1974 , yend = 13,
               lineend = "round",
               linejoin = "round",
               size = .7,
               color = my_blue,
               arrow = arrow(length = unit(0.3, "cm"),
                             type = "closed")
               )

Warning: Using `size` aesthetic for lines was deprecated in ggplot2 3.4.0.
ℹ Please use `linewidth` instead.

3.3.b.) Compared to the “historical” average number of hurricanes, was the 2012 season an above-average, a below-average, or an average season?

First, let’s create a data frame named_hurricanes which contains all of the named hurricanes between 1970 and 2021. Therefore we have

Code

# 74 mph = 64.30 knots

named_hurricanes = DF %>%
  group_by(NAME) %>%
  select(NAME, SEASON, WMO_WIND) %>%
  filter(WMO_WIND > 64.30) %>%
  count(SEASON)

Let’s display the first 5 rows of named_hurricanes

Code

head(named_hurricanes)

# A tibble: 6 × 3
# Groups:   NAME [3]
  NAME    SEASON     n
  <chr>    <int> <int>
1 AGNES     1972     6
2 ALBERTO   1982     2
3 ALBERTO   2000    47
4 ALEX      2004    13
5 ALEX      2010     8
6 ALEX      2016     9

Now using named_hurricanes we can create a new data frame num_named_hurricanes which contains all of the named hurricanes that happened per year

Code

num_named_hurricanes = named_hurricanes %>%
  group_by(SEASON) %>%
  count(SEASON)

Let’s display the first 5 rows of num_named_hurricanes

Code

head(num_named_hurricanes)

# A tibble: 6 × 2
# Groups:   SEASON [6]
  SEASON     n
   <int> <int>
1   1970     4
2   1971     6
3   1972     3
4   1973     4
5   1974     4
6   1975     6

Now we can compute the average number of named hurricanes

Code

mean(num_named_hurricanes$n)

[1] 6.480769

We can extract the total number of named hurricanes in 2012 with the following code

Code

filter(num_named_hurricanes, SEASON == 2012)

# A tibble: 1 × 2
# Groups:   SEASON [1]
  SEASON     n
   <int> <int>
1   2012    10

Therefore the average number of named hurricanes between 1970 and 2021 was about 6 per year. In 2012 there was 10 named hurricanes. Therefore 2012 was an above-average year.

We can see this in the following graph

Code

ggplot(data = num_named_hurricanes, aes(x = SEASON, y = n)) +
  geom_col(width = .8,
           fill = my_grey1) + 
  labs(title = "Total Number of Named Hurricanes",
       subtitle = "Comparing the historical average number of named hurricanes to the 2012 season",
       x = "Year",
       y = "Total") + 
  geom_col(data = filter(num_named_hurricanes, SEASON == 2012),
           color = my_red,
           fill = my_red,
           width = .8) + 
  geom_text(x = 2013.5,
            y = 10.8, 
            label = "2012",
            color = my_red,
            size = 6.8) + 
  geom_hline(yintercept = mean(num_named_hurricanes$n),
             col = my_blue,
             size = 1) + 
  geom_text(x = 1974,
            y = 10.3,
            size = 6,
            col = my_blue,
            show.legend = TRUE,
            aes(label = "Average")) +
  geom_segment(x = 1974, y = 9.5,
               xend = 1974, yend = 6.8,
               lineend = "round",
               linejoin = "round",
               size = .7,
               color = my_blue,
               arrow = arrow(length = unit(0.3, "cm"),
                             type = "closed"),
               )

3.3.c.) Compared to the “historical” average of maximum winds, was the 2012 season an above-average, a below-average, or an average season?

First, let’s create a data frame max_winds which contains the maximum winds per year. Therefore we have

Code

# For the 2021 season the wind speeds are not available

max_winds = DF %>%
  filter(SEASON != 2022) %>% 
  select(SEASON, WMO_WIND) %>%
  group_by(SEASON) %>%
  summarise(max_wmo_wind = max(WMO_WIND, na.rm = TRUE))

Let’s display the first 5 rows of max_winds

Code

head(max_winds)

# A tibble: 6 × 2
  SEASON max_wmo_wind
   <int>        <int>
1   1970          120
2   1971          140
3   1972           90
4   1973          100
5   1974          130
6   1975          120

Now we can compute the average of max winds between 1970 and 2020

Code

mean(max_winds$max_wmo_wind)

[1] 126.3462

We can extract the maximum wind speeds in 2012 with the following code

Code

filter(max_winds, SEASON == 2012)

# A tibble: 1 × 2
  SEASON max_wmo_wind
   <int>        <int>
1   2012          100

Therefore the average of maximum winds between 1970 and 2020 was about 126 knots or about 145 mph. The maximum wind speed in 2012 was about 100 knots or about 115 mph. Therefore 2012 was a below-average season with respect to maximum wind speeds.

We can show this with the following graph

Code

ggplot(data = max_winds, aes(x = SEASON, y = max_wmo_wind)) + 
  labs(title = "Maximum Wind Speeds",
       subtitle = "Comparing the historical average of maximum wind speeds to the 2012 season",
       x = "Year",
       y = "Maximum Wind in Knots") +
  geom_line(linetype = "dashed",
            color = my_grey1) +
  geom_point(size = 1.8,
             color = my_grey2) + 
  geom_point(x = 2012,
             y = 100,
             size = 3,
             color = my_red) + 
  geom_text(x = 2008.5,
            y = 96,
            color = my_red,
            size = 6.8,
            label = "2012") + 
  geom_hline(yintercept = mean(max_winds$max_wmo_wind),
             size = 1,
             color = my_blue
             ) + 
  geom_text(x = 1979,
            y = 95,
            size = 6,
            color = my_blue,
            label = "Average") + 
  geom_segment(x = 1979.8, y = 100,
               xend = 1979.8, yend = 123,
               lineend = "round",
               linejoin = "round",
               size = .7,
               color = my_blue,
               arrow = arrow(length = unit(.3, "cm"),
                             type = "closed"))

3.4. Season Summary Map

Let’s now graph map of the storm paths for 2012. First, we’ll need the following world dataset.

Code

world_df = map_data("world")

3.4.a.) Let’s now get a graph to the map of the storms

Let’s now create an additional data frame named_depressions_2012 which contains all of the individual storms and their respective time of dissipation, latitude, and longitude columns.

Code

named_depressions_2012 = depressions_2012 %>%
  select(NAME, ISO_TIME, LAT, LON) %>%
  group_by(NAME) %>%
  summarise(date = last(ISO_TIME),
            lat = last(LAT),
            lon = last(LON))

Now we can get a graph of our map

Code

ggplot(data = world_df) + 
  labs(title = "Storm Paths in 2012",
       subtitle = "The location of the names is where the storm dissipated",
       x = "Longitude",
       y = "Latitude") + 
  geom_map(aes(x = long, y = lat, map_id = region),
           map = world_df, 
           size = 1,
           fill = my_grey2) +
  xlim(-105,-1) + ylim(12,58) +
  geom_point(data = depressions_2012,
             aes(x = LON, y = LAT, color = NAME),
             size = .7,
             show.legend = FALSE) + 
  geom_label(data = named_depressions_2012,
            aes(x = lon,
                y = lat,
                label = NAME,
                color = NAME),
            show.legend = FALSE,
            size = 1.8,
            label.size = .7)

3.4.b.) Did any hurricanes hit the United States? If yes, how many, and which ones?

Similarly, let’s now create an additional data frame named_hurricanes_2012 which contains all of the individual hurricanes and their respective time of dissipation, latitude, and longitude columns. Therefore

Code

# 74 mph = 64.30 knots 

named_hurricanes_2012 = filter(depressions_2012, WMO_WIND >= 64.30) %>%
  select(NAME, ISO_TIME, LAT, LON) %>%
  group_by(NAME) %>%
  summarise(date = last(ISO_TIME),
            lat = last(LAT),
            lon = last(LON))

Code

ggplot(data = world_df) + 
  labs(title = "Storm Paths in 2012",
       subtitle = "The labeled storms are hurricanes",
       x = "Longitude",
       y = "Latitude") + 
  geom_map(aes(x = long, y = lat, map_id = region),
           map = world_df, 
           size = 1,
           fill = my_grey2) +
  xlim(-105,-1) + ylim(12,58) +
  geom_point(data = depressions_2012,
             aes(x = LON, y = LAT, color = NAME),
             size = .7,
             show.legend = FALSE) + 
  geom_label(data = named_hurricanes_2012,
            aes(x = lon,
                y = lat,
                label = NAME,
                color = NAME),
            show.legend = FALSE,
            size = 1.8,
            label.size = .7)

From our map we can see that 2 hurricanes hit the United States and they were Isaac and Sandy.