Probabilistic forecasts of the daily maximum of the Kp index produced by the SERENADE prototype model and three empirical models for the period 2010-2018

This dataset contains the probabilistic outputs of SERENADE's first prototype model dedicated to the forecasting of the \(\textit{Kp}_{\textrm{max, 24 h}}\) index for forecasting horizons ranging between 2 and 7 days. The period covered is the one of the SDOML dataset, which is 2010-05 --- 2018-12. All data is contained in a single pickle file, that can be opened in Python, using the following code lines:

\(\texttt{import pickle}\\ \texttt{with open(DATA_PATH+`/serenade_outputs.pkl', `rb') as f:}\\ ~~~~\texttt{dict_outputs = pickle.load(f)}\)

The pickle file contains a dictionnary, which itself contains Pandas DataFrames. Each DataFrame corresponds to a forecasting horizon. The dictionnary's keys are the forecasting horizons stored as strings, that is:

\(\texttt{dict_output.keys() = [`2',`3',`4',`5',`6',`7']}\)

The DataFrames are indexed by datetime. They contain the observed (true) hourly values of the daily maximum of the Kp index, the forecasts provided by SERENADE and three baseline models (Climatology model, Persistence model and 27-day Recurrence model). The forecast values include the mean and the standard deviation of the forecast normal distributions. Missing moments are due to the absence of EUV images needed to provide the forecast at the given moment.