zenodoembargoed

Skillful bias correction of offshore near-surface wind speed and wind direction forecasting based on a multi-task machine learning model

<h3>Dataset</h3> 1. observation data over 14 weather stations Variables: hourly near-surface 2-min average wind speed, wind direction  2. ECMWF-IFS forecast data over 14 weather stations Variables: hourly predictors at surface level and upper level in next 48 hours (shown in Table 1. and Table 2.) Table 1. ECMWF-IFS forecast data at surface level <div> <table> <tbody> <tr> <td> Predictors </td> <td> Abbreviation </td> <td> Unit </td> </tr> <tr> <td> Temperature at 2 m </td> <td> 2t </td> <td> ℃ </td> </tr> <tr> <td> Sea surface temperature </td> <td> sst </td> <td> ℃ </td> </tr> <tr> <td> Dewpoint temperature at 2 m </td> <td> 2d </td> <td> ℃ </td> </tr> <tr> <td> Convective precipitation in the past hour </td> <td> cp </td> <td> mm </td> </tr> <tr> <td> Mean sea level pressure </td> <td> msl </td> <td> hPa </td> </tr> <tr> <td> Zonal component of wind speed at 10 m </td> <td> 10u </td> <td> m s-1 </td> </tr> <tr> <td> Meridional component of wind speed at 10 m </td> <td> 10v </td> <td> m s-1 </td> </tr> <tr> <td> Wind speed at 10 m </td> <td> 10ws </td> <td> m s-1 </td> </tr> <tr> <td> Wind direction at 10 m </td> <td> 10wd </td> <td> ° </td> </tr> <tr> <td> Zonal component of wind speed at 100 m </td> <td> 100u </td> <td> m s-1 </td> </tr> <tr> <td> Meridional component of wind speed at 100 m </td> <td> 100v </td> <td> m s-1 </td> </tr> <tr> <td> Wind speed at 100 m </td> <td> 100ws </td> <td> m s-1 </td> </tr> <tr> <td> Wind direction at 100 m </td> <td> 100wd </td> <td> ° </td> </tr> </tbody> </table> </div> <div> </div> Table 2. ECMWF-IFS forecast data at upper level <table> <tbody> <tr> <td> Predictors </td> <td> Abbreviation </td> <td> Unit </td> </tr> <tr> <td> Relative humidity at xxx hPa </td> <td> r_Lxxx </td> <td> % </td> </tr> <tr> <td> Temperature at xxx hPa </td> <td> t_Lxxx </td> <td> ℃ </td> </tr> <tr> <td> Vertical velocity of wind at xxx hPa </td> <td> w_Lxxx </td> <td> Pa s-1 </td> </tr> <tr> <td> Zonal component of wind at xxx hPa </td> <td> u_Lxxx </td> <td> m s-1 </td> </tr> <tr> <td> Meridional component of wind at xxx hPa </td> <td> v_Lxxx </td> <td> m s-1 </td> </tr> <tr> <td> Wind speed at xxx hPa </td> <td> ws_Lxxx </td> <td> m s-1 </td> </tr> <tr> <td> Wind direction at xxx hPa </td> <td> wd_Lxxx </td> <td> ° </td> </tr> </tbody> </table> <div> </div> 3. key variables constructed by feature engineering (1) sort-term statistics, including maximum, minimum, mean and variance of key variables (2t, 10u, 10v and 10ws) from ECMWF-IFS model during the next 48 hours,  (2) long-term statistics, including mean and deviation of key variables (2t, 10u, 10v and 10ws) from ECMWF-IFS model during history 3-yr period (January 2020–December 2022),  (3) thermodynamic factors,  including the low-level wind shear between 10ws and 100ws, vertical wind shear between 200 hPa and 850 hPa, the differences between sst and 2t. <h3>Scripts</h3> 1. Random Forest model training code 2. LightGBM model training code 3. XGBoost model training code 4. TabNet-MTL model training code

ShareScore

24/100

Overall dataset sharing score

Score breakdown

These five areas show where the dataset supports — or may limit — practical reuse.

Stewardship: 12
Harmonization: 4
Access: 8
Reuse readiness: 0
Engagement: 0

Topics

wind speed wind direction forecast bias correction