Every number in this document begins as monthly temperature and rainfall from WorldClim, a global climatology built from weather-station records and interpolated to a grid. Version 2.1 carries monthly normals averaged over 1970 to 2000: average temperature, minimum and maximum temperature, precipitation, and a few other variables. We use average temperature (tavg) and precipitation (prec). This appendix shows how seasonalityr retrieves them, so any site or region in the document can be reproduced from scratch, and so you can point the same tools at places of your own.
get_climate() does the whole job: it downloads the WorldClim rasters once, caches them, and extracts the twelve monthly values at each site’s coordinates. The input is the part you supply yourself, and it is just a small table with one row per site and four columns. name is a label for the site. zone is your own grouping label, carried straight through to the output, where it is handy later for coloring or faceting. lat and lon are the coordinates in decimal degrees, with south and west written as negative numbers.
To show how to bring your own sites, here are two that are nowhere in the package’s bundled cities: Honolulu, where this is being written, and Reykjavik, near the cold end of the seasonality range. You build the table by hand. That is the whole trick: once it has those four columns, get_climate() will take it.
## a table of your own sites: one row each, the four required columns.
## decimal degrees; south and west are negative.
my_sites <- read_csv(
"name, zone, lat, lon
Honolulu, tropical, 21.31, -157.86
Reykjavik, subpolar, 64.13, -21.90",
trim_ws = TRUE)
my_sites |>
gt() |>
tab_header(title = "Your input: a site table you build by hand")| Your input: a site table you build by hand | |||
| name | zone | lat | lon |
|---|---|---|---|
| Honolulu | tropical | 21.31 | -157.86 |
| Reykjavik | subpolar | 64.13 | -21.90 |
Then hand that table to get_climate(). The call is identical whether the sites are your own or the package’s:
## download once, cache, and extract the monthly normals at your sites
clim <- get_climate(my_sites, res = 10)
clim
#> # A tibble: 24 x 5
#> name zone month tavg prec
#> <chr> <chr> <int> <dbl> <dbl>
#> 1 Honolulu tropical 1 22.6 139
#> 2 Honolulu tropical 2 22.7 78
#> 3 Honolulu tropical 3 23.4 96
#> ...The result has one row per site-month, carrying your name and zone through unchanged, ready for compute_indices() or indices_table().
The first call to get_climate() downloads the global WorldClim layers through the geodata package. This is the one heavy step. At the default 10 arc-minute resolution the twelve temperature and twelve precipitation layers together are a modest download; at finer resolutions they grow quickly. The data is written under the path directory and reused on every later call, so the download happens only once.
Leaving path at its default puts the cache in the session’s temporary directory, which is cleared when R restarts. For repeated work, give it a real folder. The rasters land in wc_cache/climate/wc2.1_10m/, and geodata will find them there next time rather than downloading again.
WorldClim ships at several grid resolutions, named by the angular size of a cell: 10, 5, and 2.5 arc-minutes, and 30 arc-seconds. The res argument selects among them. Coarser grids are smaller and faster and are plenty for points spread across a continent; finer grids resolve coastlines, valleys, and rain shadows, at the cost of much larger downloads. The point work and the regional maps in this document use 10 arc-minutes throughout, which keeps the data manageable while still separating places a few tens of kilometers apart.
This document uses only average temperature and precipitation, the two channels the climate diagram is built from. WorldClim 2.1 also serves monthly minimum and maximum temperature (tmin, tmax), along with solar radiation, wind, and water-vapor pressure. The minimum-temperature layers are what a true Walter-Lieth frost line needs, the marker for months that can freeze. Our warm-desert examples never use it, but a temperate or high-latitude site, the kind the document points toward in its closing chapter, could retrieve tmin the same way and add the frost story back.
To keep the main chapters light, the desert trio’s monthly normals were retrieved once and saved to a small file, data/desert_trio_monthly.csv, which Chapter 2 reads. That file was produced by the recipe below: retrieve, add a readable desert label and the moisture value m = P - 2T, round, and write.
library(seasonalityr)
library(dplyr)
library(readr)
trio <- dplyr::filter(cities, name %in% c("Las Vegas", "Tucson", "El Paso"))
clim <- get_climate(trio, res = 10, path = "wc_cache")
trio_csv <- clim |>
mutate(desert = recode(name,
"Las Vegas" = "Mojave (Las Vegas)",
"Tucson" = "Sonoran (Tucson)",
"El Paso" = "Chihuahuan (El Paso)"),
tavg = round(tavg, 2),
prec = round(prec),
m = round(prec - (2 * tavg), 2)) |>
select(name, desert, month, tavg, prec, m)
write_csv(trio_csv, "data/desert_trio_monthly.csv")The same pattern produces the data behind any figure in the document. Pick sites, call get_climate(), compute the indices, and the rest follows.