In timtrice/Hurricanes: Web Scraper for Atlantic and East Pacific Hurricanes and Tropical Storms

library(dplyr)
library(ggplot2)
library(rrricanes)
library(sf)

Most storms will contain a variation of GIS datasets that can be plotted with ggplot2. The helper functions for this have the prefix 'gis'.

All products are experimental and there maybe fluctuations particularly in current datasets.

In general, datasets are available for storms dated back to 1998. However, products such as Wind Speed Probabilities only go back to 1999.

Some datasets require the use of the storm key and an optional advisory. Other products require a datetime value and cannot be isolated by storm key or advisory. The datetime values are not based on the issue time of the advisory, but rather three hours prior. For example, if you are seeking a dataset where the forecast/advisory was issued at 9:00AM UTC, you will want the dataset for 6:00AM UTC. This will be explained a little further below.

Build a Tracking Chart

There are three functions available to help you plot GIS data; tracking_chart, al_tracking_chart and ep_tracking_chart. al_tracking_chart and ep_tracking_chart are just helpers centered on the Atlantic and northeast Pacific ocean, respectively.

args(rrricanes::tracking_chart)

The countries and states parameters are TRUE by default. This means a basic call to tracking_chart will return a map with country and state borders. The res parameter is resolution; one of 110, 50 or 10 nautical miles. Resolutions 110nm and 50nm can be used immediately. To use lower resolution you must install the rnaturalearthdatahires package from ropensci:

# this should not be needed since package is required
install.packages("rnaturalearthhires",
                 repos = "http://packages.ropensci.org",
                 type = "source")

tracking_chart will print a basic tracking chart (a map of the planet).

tracking_chart()

You can pass typical aes parameters to refine the color and fill of the plot; remember the tracking chart is a ggplot object.

tracking_chart(color = "black", fill = "white", size = 0.1)

You may choose to only show coastline data instead. In this case, just set the countries parameter to FALSE.

tracking_chart(countries = FALSE, res = 50, color = "black", fill = "white", 
               size = 0.1)

For the purposes of this vignette we'll focus on Atlantic storms.

(p <- al_tracking_chart(color = "black",
                        fill = "white", 
                        size = 0.1,
                        res = 50))

The res parameter defines the resolution of the chart presented. Options are in 110nm, 50nm and 10nm. The lower the resolution the longer the chart takes to be built.

States cannot be drawn on resolution greater than 50nm.

GIS Datasets

There are several datasets that are published for active cyclones. The following functions are designed to return the URL to those datasets:

gis_advisory
gis_prob_storm_surge
gis_windfield
gis_latest

Advisory Package

gis_advisory(key = "AL182012", advisory = "18")

# Needs to update to support sf
df.gis_adv <- gis_advisory(key = "AL182012",
                            advisory = "18") |>
                            gis_download()

names(df.gis_adv)

For this particular storm and advisory, included are the base line, point and polygon datasets along with a dataset for watches and warnings. The objects returned are spatial dataframes contained within the list of dataframes, df.gis_adv.

Line track

# still in sp
str(df.gis_adv$al182012.018_5day_lin)

As we're dealing with a SpatialLinesDataFrame which needs to be modified, you can use the helper function shp_to_df for plotting.

fcst_line <- 
  shp_to_df(df.gis_adv$al182012.018_5day_lin)

p + geom_path(data = fcst_line, 
              aes(long, lat, group = FCSTPRD))

There are two groups of data in the set: one for 72-hour forecast period and one for 120-hour. Grouping by FCSTPRD will show the forecast track correctly.

There is a pretty wide field of view on the map above. You can use sp::bbox to "zoom in" on the map.

(bb <- sf::st_bbox(df.gis_adv$al182012.018_5day_lin))

(p2 <- p + geom_path(data = fcst_line, aes(long, lat, group = FCSTPRD)) + 
    coord_equal(xlim = c(bb[1] - 5, bb[3] + 5), 
                ylim = c(bb[2] - 5, bb[4] + 5)))

Point track

Each forecast position is also included in the points dataframe. You can access this in the df.gis_adv$al182012.018_5day_pts@data object; no conversion necessary.

(p3 <- p2 + 
   geom_point(data = df.gis_adv$al182012.018_5day_pts@data,
              aes(LON, LAT)))

Forecast Cone

Forecast cone data is contained in the polygon dataset. To deal with this dataset you can use the shp_to_df function again or take the slightly longer way:

fcst_cone <- df.gis_adv$al182012.018_5day_pgn
fcst_cone@data$id <- rownames(fcst_cone@data)
fcst_cone.points <- broom::tidy(fcst_cone, region = "id")
fcst_cone <- dplyr::left_join(fcst_cone.points, fcst_cone@data, by = "id")

p3 + geom_polygon(data = fcst_cone |> filter(FCSTPRD == 120), 
                 aes(long, lat, group = group, fill = factor(FCSTPRD)), 
                 alpha = 0.5) + 
    geom_polygon(data = fcst_cone |> filter(FCSTPRD == 72), 
                 aes(long, lat, group = group, fill = factor(FCSTPRD)), 
                 alpha = 0.5)

Note that in some GIS packages the forecast cones may be identical (though they shouldn't be). I've noticed it with Hurricanes Ike and Matthew; it's in the raw dataset.

Watches and Warnings

You can also plot watches and warnings if any are in effect for the package release time.

ww_line <- shp_to_df(df.gis_adv$al182012.018_ww_wwlin)

p3 + geom_path(data = ww_line, 
               aes(x = long, y = lat, group = group, color = TCWW), size = 1)

In the example above you can see tropical storm warnings issued for the Bahamas, North Carolina and portions of the South Carolina and Florida coast while tropical storm watches are in effect for northern Florida and South Carolina.

Probabilistic Storm Surge

The Tropical Cyclone Storm Surge Probabilities data shows the probability, in percent, of a specified storm surge occurring during the forecast period indicated. The product is based upon an ensemble of Sea, Lake, and Overland Surge from Hurricanes (SLOSH) model runs using the National Hurricane Center (NHC) official advisory and accounts for track, size, and intensity errors based on historical errors.

gis_prob_storm_surge(key = "AL142016", 
                  products = list(psurge = 0), 
                     datetime = "20161006")

# Make this work with sf
df.gis_storm_surge <- 
  gis_prob_storm_surge(key = "AL142016", 
                   products = list(psurge = 0), 
                  datetime = "20161006") |> 
  last() |> 
  gis_download()

prob_surge <- shp_to_df(df.gis_storm_surge$al142016_2016100618_gt0)
bb <- sf::st_bb(df.gis_storm_surge$al142016_2016100618_gt0)

p + geom_path(data = prob_surge, 
               aes(x = long, y = lat, 
                   group = group, 
                   color = PSurge00c), 
              size = 1, alpha = 0.5) + 
    coord_equal(xlim = c(bb[1,1], bb[1,2]),
                ylim = c(bb[2,1], bb[2,2]))

Current and Forecast Wind Field Radii

Wind radii data may also be available for some cyclone advisory packages. This is the radius to which a minimum sustained wind speed may be felt from the center of circulation.

gis_windfield("AL142016", advisory = "33")

df.gis_wind_radii <- gis_windfield("AL142016", advisory = "33") |> 
                     gis_download()

wf_init <- shp_to_df(df.gis_wind_radii$al142016_2016100606_initialradii)
bb <- sp::bbox(df.gis_wind_radii$al142016_2016100606_forecastradii)

(p4 <- p + geom_polygon(data = wf_init, 
                  aes(x = long, y = lat, fill = factor(RADII)), alpha = 0.5) + 
    coord_equal(xlim = c(bb[1,1], bb[1,2]), ylim = c(bb[2,1], bb[2,2])))

Additionally, forecast wind radii data is also generally available in some packages

wf_fcst <- shp_to_df(df.gis_wind_radii$al142016_2016100606_forecastradii)

p4 + geom_polygon(data = wf_fcst, 
                  aes(x = long, y = lat, group = group, fill = factor(RADII)), 
                  alpha = 0.5)

Wind Speed Probabilities

Wind Speed probabilities show the chance of experiencing a minimum-sustained winds of 34, 50 and 64 knots with a given period of time (typically, 120 hours). These products are not storm-specific but are global so other active cyclones in other basins may also appear.

#returning a long vector of NA
gis_wsp(datetime = "2016100606", res = 0.5)

df.gis_wsp <- 
  gis_download(
    "https://www.nhc.noaa.gov/gis/forecast/archive/2016100606_ws
    120hrhalfDeg.zip"
  )

Cumulative Probability for >34kt Winds

bb <- sf::st_bbox(df.gis_wsp$`2016100606_wsp34knt120hr_halfDeg`)
p + 
  geom_sf(
    data = st_as_sf(df.gis_wsp$`2016100606_wsp34knt120hr_halfDeg`), 
    aes(color = PWIND120)
  ) +
  coord_sf(xlim = c(bb[1,1], bb[1,2]), ylim = c(bb[2,1], bb[2,2]))

Cumulative wind speed probability for >50kt winds:

bb <- sp::bbox(df.gis_wsp$`2016100606_wsp50knt120hr_halfDeg`)
p + 
  geom_sf(
    data = st_as_sf(df.gis_wsp$`2016100606_wsp50knt120hr_halfDeg`), 
    aes(color = PWIND120)
  ) +
  coord_sf(xlim = c(bb[1,1], bb[1,2]), ylim = c(bb[2,1], bb[2,2]))

Cumulative probability for >64kt winds:

bb <- sp::bbox(df.gis_wsp$`2016100606_wsp64knt120hr_halfDeg`)
p + 
  geom_sf(
    data = st_as_sf(df.gis_wsp$`2016100606_wsp64knt120hr_halfDeg`), 
    aes(color = PWIND120)
  ) +
  coord_sf(xlim = c(bb[1], bb[3]), ylim = c(bb[2], bb[4]))