ISSN 1000-0526
CN 11-2282/P
CN 11-2282/P
Multi-site wind speed and direction forecasting based on two-step bias correction
- Article
- Figures
- Metrics
- Preview PDF
- Reference
- Related
- Cited by
- Materials
Abstract:
To enhance the forecast accuracy of average wind speed and direction, this study utilized the 2-minute average wind speed and direction data from ground observations provided by the China Meteorological Administration (CMA) from January 2021 to December 2022, in conjunction with the 24-hour forecasted 10-meter wind field data from the deterministic model of the European Centre for Medium-Range Weather Forecasts (ECMWF). A two-step bias correction method tailored for multi-site wind speed and direction forecasts was developed for 662 national observation stations located within the range of 110°E-120°E and 30°N-40°N. Given that wind direction is a circular variable, traditional statistical post-processing methods are challenging to directly apply for wind direction correction. Therefore, this study adopted the u and v wind components, which simultaneously contain wind speed and direction information, as key variables. Considering the significant differences in u and v wind forecast errors among stations, a site-specific modeling strategy was employed to enhance the relevance and precision of the model. Although the joint distribution of u and v winds approximates a two-dimensional Gaussian distribution, and joint modeling could potentially improve wind speed and direction forecasts, it may also intertwine wind speed and direction forecast errors, increasing model instability. Consequently, this study constructed separate univariate regression correction models for the u and v winds at each station. After correcting the u and v winds, the wind direction forecast errors decreased at most stations. However, when the numerical model forecasted excessively small u and v winds, a sign reversal occurred after correction, primarily in cases of low forecasted wind speeds. Regarding wind speed forecasts, despite the u and v wind corrections, deficiencies still existed in forecasting stronger winds. To address this, this study utilized the quantile matching method to further correct the wind speeds synthesized from the corrected u and v winds, referred to as the two-step bias correction method. To further validate the effectiveness of the two-step bias correction method in practical operational applications, an independent sample test was conducted using data from January to December 2023. In the comparison of wind direction forecasts, the two-step bias correction method demonstrated good generalization ability and forecast performance. Among various wind speed forecast comparisons, this method exhibited the smallest mean error and root mean square error (RMSE) for wind speeds of 10 m/s and above. Additionally, in cases of strong winds such as those associated with cold fronts and typhoons, the method also showed significant forecast advantages. For forecasts of all wind speeds and those specifically at 10 m/s and above, the RMSE after two-step bias correction was reduced by 18.6% and 29.6%, respectively, compared to the uncorrected numerical model.
Keywords:
Project Supported:
Clc Number:
Mobile website
