{"id":3615,"date":"2015-08-20T11:41:50","date_gmt":"2015-08-20T16:41:50","guid":{"rendered":"http:\/\/rud.is\/b\/?p=3615"},"modified":"2018-03-07T16:43:28","modified_gmt":"2018-03-07T21:43:28","slug":"track-hurricane-danny-with-r-leaflet","status":"publish","type":"post","link":"https:\/\/rud.is\/b\/2015\/08\/20\/track-hurricane-danny-with-r-leaflet\/","title":{"rendered":"Track Hurricane Danny (Interactively) with R + leaflet"},"content":{"rendered":"

\"poster<\/p>\n

Danny became the [first hurricane of the 2015 Season](http:\/\/www.accuweather.com\/en\/weather-news\/atlantic-gives-birth-to-tropical-depression-four-danny\/51857239), so it’s a good time to revisit how one might be able to track them with R.<\/p>\n

We’ll pull track data from [Unisys](http:\/\/weather.unisys.com\/hurricane\/atlantic\/2015\/index.php) and just look at Danny, but it should be easy to extrapolate from the code.<\/p>\n

For this visualization, we’ll use [leaflet](http:\/\/rstudio.github.io\/leaflet\/) since it’s all the rage and makes the plots interactive without any real work (thanks to the very real work by the HTML Widgets folks and the Leaflet.JS folks).<\/p>\n

Let’s get the library calls out of the way:<\/p>\n

library(leaflet)\r\nlibrary(stringi)\r\nlibrary(htmltools)\r\nlibrary(RColorBrewer)<\/pre>\n
Now, we’ll get the tracks:<\/p>\n
danny <- readLines(\"http:\/\/weather.unisys.com\/hurricane\/atlantic\/2015\/DANNY\/track.dat\")<\/pre>\n
Why aren't we using `read.csv` or `read.table` directly, you ask? Well, the data is in a _really_ ugly format thanks to the spaces in the `STATUS` column and two prefix lines:<\/p>\n
Date: 18-20 AUG 2015\r\nHurricane-1 DANNY\r\nADV  LAT    LON      TIME     WIND  PR  STAT\r\n  1  10.60  -36.50 08\/18\/15Z   30  1009 TROPICAL DEPRESSION\r\n  2  10.90  -37.50 08\/18\/21Z    -     - TROPICAL DEPRESSION\r\n  3  11.20  -38.80 08\/19\/03Z    -     - TROPICAL DEPRESSION\r\n  4  11.30  -40.20 08\/19\/09Z    -     - TROPICAL DEPRESSION\r\n  5  11.20  -41.10 08\/19\/15Z    -     - TROPICAL DEPRESSION\r\n  6  11.50  -42.00 08\/19\/21Z    -     - TROPICAL DEPRESSION\r\n  7  12.10  -42.70 08\/20\/03Z    -     - TROPICAL DEPRESSION\r\n  8  12.20  -43.70 08\/20\/09Z    -     - TROPICAL DEPRESSION\r\n  9  12.50  -44.80 08\/20\/15Z    -     - TROPICAL DEPRESSION\r\n+12  13.10  -46.00 08\/21\/00Z   70     - HURRICANE-1\r\n+24  14.00  -47.60 08\/21\/12Z   75     - HURRICANE-1\r\n+36  14.70  -49.40 08\/22\/00Z   75     - HURRICANE-1\r\n+48  15.20  -51.50 08\/22\/12Z   70     - HURRICANE-1\r\n+72  16.00  -56.40 08\/23\/12Z   65     - HURRICANE-1\r\n+96  16.90  -61.70 08\/24\/12Z   65     - HURRICANE-1\r\n+120  18.00  -66.60 08\/25\/12Z   55     - TROPICAL STORM<\/pre>\n
But, we can put that into shape pretty easily, using `gsub` to make it easier to read everything with `read.table` and we just skip over the first two lines (we'd use them if we were doing other things with more of the data).<\/p>\n
danny_dat <- read.table(textConnection(gsub(\"TROPICAL \", \"TROPICAL_\", danny[3:length(danny)])), \r\n           header=TRUE, stringsAsFactors=FALSE)<\/pre>\n
Now, let's make the data a bit prettier to work with:<\/p>\n
# make storm type names prettier\r\ndanny_dat$STAT <- stri_trans_totitle(gsub(\"_\", \" \", danny_dat$STAT))\r\n\r\n# make column names prettier\r\ncolnames(danny_dat) <- c(\"advisory\", \"lat\", \"lon\", \"time\", \"wind_speed\", \"pressure\", \"status\")<\/pre>\n
Those steps weren't absolutely necessary, but why do something half-baked (unless it's chocolate chip cookies)?<\/p>\n
Let's pick better colors than Unisys did. We'll use a color-blind safe palette from Color Brewer:<\/p>\n
danny_dat$color <- as.character(factor(danny_dat$status, \r\n                          levels=c(\"Tropical Depression\", \"Tropical Storm\",\r\n                                   \"Hurricane-1\", \"Hurricane-2\", \"Hurricane-3\",\r\n                                   \"Hurricane-4\", \"Hurricane-5\"),\r\n                          labels=rev(brewer.pal(7, \"YlOrBr\"))))<\/pre>\n
And, now for the map! We'll make lines for the path that was already traced by Danny, then make interactive points for the forecast locations from the advisory data:<\/p>\n
last_advisory <- tail(which(grepl(\"^[[:digit:]]+$\", danny_dat$advisory)), 1)\r\n\r\n# draw the map\r\nleaflet() %>% \r\n  addTiles() %>% \r\n  addPolylines(data=danny_dat[1:last_advisory,], ~lon, ~lat, color=~color) -> tmp_map\r\n\r\nif (last_advisory < nrow(danny_dat)) {\r\n  \r\n   tmp_map <- tmp_map %>% \r\n     addCircles(data=danny_dat[last_advisory:nrow(danny_dat),], ~lon, ~lat, color=~color, fill=~color, radius=25000,\r\n             popup=~sprintf(\"Advisory forecast for +%sh (%s)<\/b>
\r\n                           Position: %3.2f, %3.2f\r\n                           Expected strength: %s<\/strong><\/span>\r\n                           Forecast wind: %s (knots)Forecast pressure: %s\",\r\n                           htmlEscape(advisory), htmlEscape(time), htmlEscape(lon),\r\n                           htmlEscape(lat), htmlEscape(color), htmlEscape(status), \r\n                           htmlEscape(wind_speed), htmlEscape(pressure)))\r\n}\r\n\r\nhtml_print(tmp_map)<\/pre>\n