Why a station has decades of history on its first day
A station reports for the first time and the platform has exactly 1 reading to work with. But an anomaly advisory needs a normal to compare against, and a normal needs decades, not a single reading, so this dataset supplies them up front: the moment a station reports its position, the platform backfills daily climate for that point from a reanalysis reaching to 1940 and a rainfall record reaching to 1981, then keeps the series topped up weekly from there.
Fields of the climate archive
7 fields land here for every day: maximum, minimum and mean temperature, rainfall, maximum wind, radiation, and reference evapotranspiration. The source tag separates the 2 underlying archives from each other. The reanalysis is a model’s own reconstruction of the atmosphere on a 9 km grid, so it was never going to match a station’s thermometer to the decimal; what it matches is the station’s climatology, which is exactly the job it exists to do. The station’s own daily means sit inside the series too, as its first measured years, and the gap between the 2 over a year is simply the station’s local offset from the grid, typically a degree or 2 at altitude.
Uses on the platform
The frost advisory reads a station’s own minimum against the location’s minimum climatology. The vegetation-anomaly advisory needs season rainfall measured against normal for the location, which this dataset can give before the station has earned a season of its own. Reference evapotranspiration across decades is simply the location’s long-run water demand, the number any irrigation design has to start from.
Reading a long series
A decade of daily rows works out to 3,650 rows per field. Pass every=30d with agg=mean for a monthly chart, or every=365d with agg=sum on rainfall for annual totals instead.