This R package provides a few convenience functions to obtain weather data (specifically, daily and 3-hourly forecasts, and current observations) for a specific location from the Australian Government Bureau of Meteorology (BOM) through the BOM's REST API. Sadly, the BOM has recently taken a more restrictive position towards providing data access by means of automated processes. Currently, only limited data are available through the BOM's API endpoint https://api.weather.bom.gov.au/v1/. I found the documentation provided in BoM Weather Docs invaluable for implementing these functions.
This project is neither related to nor endorsed by the BOM. With (continued and expected) changes to how the BOM manages (read: restricts) automatic access to data, some or all of the functions might break at any time.
All data provided by the BOM are under copyright.
The easiest way to install the R package is from GitHub via remotes.
remotes::install_github("mevers/weatherBOM")We start by loading the library and looking up the location ID for Canberra Airport.
library(weatherBOM)
bom_search_location("Canberra")This returns a data.frame
geohash id name postcode state
1 r3dp5hh Canberra-r3dp5hh Canberra 2609 ACT
2 r3dp4vp Canberra Airport-r3dp4vp Canberra Airport 2609 ACT
from which we can read off the location ID in column geohash. For Canberra Airport, this is r3dp4vp.
We can now use the location ID to get 3-hourly forecasts. To make things more interesting, we visualise temperature forecasts and also include labels denoting the predicted wind speed & direction as well as probability of rain.
library(tidyverse)
bom_forecasts("r3dp5hh", "3-hourly") %>%
mutate(wind.angle = compass2angle(wind.direction)) %>%
ggplot(aes(time, temp)) +
geom_line() +
geom_point(shape = 21, size = 8, colour = "white", fill = "white") +
geom_text(
aes(y = 1, label = sprintf(
"%s°C\n%s km/h %s\nRain %s%%",
temp, wind.speed_kilometre, wind.direction, rain.chance)),
size = 2.5, family = "Helvetica") +
geom_text(aes(angle = 360 - wind.angle), size = 4, label = "↓") +
labs(x = "", y = "Temperature [in °C]") +
scale_x_datetime(date_breaks = "3 hour", date_labels = "%a %d\n%H:%M") +
expand_limits(y = 0) +
theme_bw()
ggsave("example_3-hourly.png", width = 12, height = 6)
To show daily minimum and maximum temperature forecasts we can use bom_forecasts with implicit (default) freq = "daily". The following plot shows +7 days maximum and minimum temperatures in red and blue, respectively.
library(tidyverse)
bom_forecasts("r3dp4vp") %>%
pivot_longer(starts_with("temp_")) %>%
ggplot(aes(date, value, colour = name)) +
geom_line() +
geom_point(shape = 21, size = 10, colour = "white", fill = "white") +
geom_text(aes(label = sprintf("%s°C", value)), size = 3) +
scale_color_manual(values = c("red", "blue")) +
guides(colour = "none") +
labs(x = "", y = "Temperature [in °C]") +
scale_x_datetime(date_breaks = "1 day", date_labels = "%a, %d %b") +
expand_limits(y = 0) +
theme_minimal()
ggsave("example_daily.png", width = 12, height = 6)
As an (educational) exercise, we use prophet to fit a seasonal model that 3-hourly forecasts and then use the model to obtain hourly "interpolated" values.
library(prophet)
m <- bom_forecasts("r3dp5hh", "3-hourly") %>%
select(ds = time, y = temp) %>%
prophet(
growth = "linear",
n.changepoints = 0,
daily.seasonality = TRUE,
weekly.seasonality = FALSE,
yearly.seasonality = FALSE)
interpolate <- data.frame(
ds = seq.POSIXt(min(m$history$ds), max(m$history$ds), by = "1 hour"))
forecast <- predict(m, interpolate)
plot(m, forecast) +
labs(x = "", y = "Temperature [in °C]") +
scale_x_datetime(date_breaks = "3 hour", date_labels = "%a %d\n%H:%M") +
expand_limits(y = 0) +
theme_bw()
ggsave("example_interpolate.png", width = 12, height = 6)