A simple prototype R wrapper to access past weather forecasts from the Global Forecast System (GFS) data as housed in the National Digital Forecast System archive.
To use the wrapper, install ncdf4 and rvest.
Rather than haphazardly searching for data series codes to access GFS from a spreadsheet, simply use translateFields to get a dataframe with the necessary information. Choose from "snow","quant_precip", "tmin", "tmax", "skycover", "prob_precip", "wind_gust", "wind_speed", "rel_humidity","sig_wave_hgt", "dew_point"
translateFields(vector)
Returns a dataframe of gridded forecasts. Inputs required:
getForecast(series, yyyymmdd, vintages = c(1,13), milestone = "00")
- series = is an object of data series as obtained from
translateFields() - yyyymmdd = a string value containing year, month and day for the day a model with run (not the forecast target date)
- vintages = Determines which model run to extract from the Thredds server as there there are 24 forecasts per day. Defaults to the 1st and 13th forecasts
- milestone = character value denoting desired hour for which a forecast is produced. Default is "00". For example: If the specified yyyymmdd is 9/10, then "00" would return 12am on 9/11. There is an exception: if the series is tmin, getForecast() will return 12pm on 9/11 is returned due to the forecasting cycles
Fields are named based on the series name + reference hour (hours since) + time step since ref hour.
(Support function) For a given day, obtain a list of all available files in the Thredds server. Input string containing Year, month and day (8 characters).
getThreddsDir(yyyymmdd)
Get series codes using translateFields()
multi_series <- translateFields(c("snow","tmin","tmax", "quant_precip", "prob_precip"))
Now obtain a dataframe of forecasts produced on September 2, 2010 with a target of 0-hour on September 3
test <- getForecast(series = multi_series, yyyymmdd = "20100902", vintages = c(1, 13), milestone = "00")