Description Usage Format Details Source References
This is a dataset containing the post-processed forecasts of five ensemble weather prediction models for two-meter temperature from March, 2019 to March, 2021.
1 |
A named list with five entries, each containing a data.table
with 731 rows.
initial condition date (with format YYYY-MM-DD)
latitude of the forecast location (in degrees)
longitude of the forecast location (in degrees)
forecast in Kelvins sampled from the gaussian distribution. The forecasts are sorted in increasing order.
the measured two-meter air temperature in Kelvins
Each ensemble has been post-processed with the non homogeneous regression technique, described in Gneiting et al.(2005). In a nutshell the true distribution is supposed to be gaussian, with mean and standard deviation being a linear function of the ensemble mean and standard deviation (respectively). The intercept and slope of each regression is determined by minimizing the CRPS over a 60-day sliding window. The forecast in the data set is a sample of 30 values from this gaussian distribution.
The five ensemble models are named based on the raw ensemble: CWAO (from ECCC), DEMS (from NCMRWF), ECMF (from ECMWF), EGRR (from UKMO) and RKSL (from KMA).
The raw forecasts are the ones at the nearest grid point to Toulouse-Blagnac station (France) in the TIGGE data set. The observation is the two-meter height temperature measured at this same station, at 06UTC. The forecast initial time is 00UTC, with a 30 hour lead-time.
https://apps.ecmwf.int/datasets/data/tigge/levtype=sfc/type=pf/
https://donneespubliques.meteofrance.fr/?fond=produit&id_produit=91&id_rubrique=32
Gneiting, Tilmann, et al. "Calibrated probabilistic forecasting using ensemble model output statistics and minimum CRPS estimation." Monthly Weather Review 133.5 (2005): 1098-1118. doi:https://doi.org/10.1175/MWR2904.1
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.