OpenForecast results in 2020-2021

<p><a href="https://openforecast.github.io">OpenForecast</a> is the first national-scale operational runoff forecasting system in Russia. The presented data supports a research article on a long-term assessment of OpenForecast performance in 2020-2021.</p> <p>File listing:</p> <ul> <li>calibration_vs_hindcast.npy -- Python dictionary that provides results of efficiency assessment for calibration and evaluation (hindcast) periods in terms of NSE and KGE metrics for individual hydrological models (GR4J_NSE, GR4J_KGE, HBV_NSE, HBV_KGE).</li> <li> <p>hindcast_vs_forecast.npy -- Python dictionary that provides results of efficiency assessment for hindcast, pre-operational hindcast, and forecast periods in terms of NSE and KGE metrics for individual hydrological models (GR4J_NSE, GR4J_KGE, HBV_NSE, HBV_KGE), as well as their ensemble mean (ENS).</p> </li> <li> <p>meteo_forecast.npy -- Python dictionary that reports correlation coefficients between ICON and ERA5 reanalysis for air temperature and precipitation forecasts.</p> </li> <li> <p>users.csv -- daily numbers of OpenForecast users.</p> </li> </ul> <p>Sample code for data access:</p> <pre><code class="language-python">import numpy as np import pandas as pd calibration_hindcast = np.load("calibration_vs_hindcast.npy", allow_pickle=True).item() hindcast_forecast = np.load("hindcast_vs_forecast.npy", allow_pickle=True).item() meteo_forecasts = np.load("meteo_forecast.npy", allow_pickle=True).item() users = pd.read_csv("users.csv", index_col=0, parse_dates=True, dayfirst=True) % pandas dataframe for the GR4J_KGE model efficiency in terms of NSE for calibration and hindcast periods calibration_hindcast["GR4J_KGE"]["NSE"] % pandas dataframe for the ensemble mean efficiency in terms of NSE for hindcast and forecast periods hindcast_forecast["ENS"]["NSE"] % pandas dataframe for correlation coefficients between ICON and ERA5 for precipitation forecasts meteo_forecasts["P"]["Correlation"] % available keys of Python dictionaries could be checked as follows calibration_hindcast.keys()</code></pre> <p>&nbsp;</p>