rmet

rmet downloads and parses historical hourly weather data from the automatic stations of Brazil's National Institute of Meteorology (INMET). It handles the format and encoding changes across years, validates and resumes downloads, keeps a local cache up to date, and builds daily summaries in local time.

Upgrading from rmet 0.1.0 or 0.2.0? The station catalogue bundled with those versions was incorrect (most station codes pointed to the wrong station). Please upgrade and re-check any station metadata obtained offline or with source = "bundled". See NEWS for all corrections.

Installation

install.packages("rmet")

# development version
# install.packages("pak")
# pak::pak("rodrigosqrt3/rmet")
install.packages("rmet", repos = c('https://rodrigosqrt3.r-universe.dev', 'https://cloud.r-project.org'))

Quick start

library(rmet)

# stations near Porto Alegre
inmet_nearest(latitude = -30.03, longitude = -51.23, n = 3)

# hourly data (archives are downloaded once and cached)
weather <- inmet_get(years = 2022:2024, stations = "A801")

# flag implausible values, check completeness, summarise per local day
weather <- inmet_qc(weather)
inmet_completeness(weather)
daily <- inmet_daily(weather)

Timestamps are returned in UTC, as in the source files, and mark the end of the hour each row describes. inmet_daily() assigns hours to each station's local day, skips values flagged by inmet_qc(), and only reports a daily value when enough hours are available.

Main functions

Function	Purpose
`inmet_get()`	Download (if needed) and read hourly data in one call.
`inmet_daily()`	Daily summaries per station and local day.
`inmet_qc()`	Flag implausible hourly values.
`inmet_completeness()`	Percentage of valid hours per station and month or year.
`inmet_stations()`	Search the official station catalogue.
`inmet_nearest()`	Find the stations closest to a location.
`inmet_timezone()`	Local time zone of Brazilian stations.
`inmet_as_sf()`	Convert stations or data to `sf` points for maps.
`inmet_download()`	Download, validate and update annual ZIP archives.
`inmet_read()`	Read cached archives into a data frame.
`inmet_extract()`	Extract the original CSV files.
`inmet_cache_status()`, `inmet_cache_clear()`	Manage the cache.

How it works

Catalogue. inmet_stations() uses INMET's station API, cached for 30 days, and falls back to a dated snapshot of the official catalogue shipped with the package.
Downloads. INMET publishes one archive per year (about 100 MB for recent years) with every station. Downloads go to a .part file, are checked against the size reported by the server and resumed after interruptions. Archives of the current and previous year are re-downloaded when INMET publishes a newer version.
Parsing. Files are read directly from the archives. Date and hour formats, Latin-1 and UTF-8 encodings, decimal commas, -9999 sentinels and damaged header text from different years are handled; columns get stable English names (see ?inmet_read).

Data quality

INMET describes automatic-station observations as raw data that have not undergone consistency validation. inmet_qc() flags gross errors (impossible values, inconsistencies, spikes, stuck sensors), but it cannot certify that every measurement is meteorologically plausible. See vignette("daily-data", package = "rmet") for time-zone, missing-data and quality-control advice.

Data source and license

Data are retrieved from INMET and are not redistributed by rmet, apart from the station catalogue snapshot and a one-station example dataset. Package code is licensed under GPL (>= 3). Please report problems at https://github.com/rodrigosqrt3/rmet/issues.