Extended-range Forecast Skill and Source Attribution of Daily Wind Power Density over Tropical Lands

Accurate extended-range forecasting of wind power is important in modern energy systems. As the global share of wind energy in the power grid continues to rise, uncertainties in wind power generation pose significant challenges to grid stability, power dispatch, and energy trading. Extended-range forecasts (1–4 weeks ahead) enable grid operators to optimize power generation schedules, reduce reserve requirements, and minimize integration costs.

The subseasonal-to-seasonal (S2S) predictability of wind power density (WPD) over tropical land regions at 1–4-week lead times was investigated using observational data (Global Wind Atlas), reanalysis products (ERA5), and S2S model outputs (ECMWF). Diagnostic analysis reveals that ERA reanalysis systematically underestimates daily mean wind speeds across global land areas (global mean bias is −1.06 m/s), with larger discrepancies at higher altitude regions.

After ERA-based wind speed bias correction, the prediction skill of WPD was assessed based on correlation coefficient (Cor). ECMWF S2S models exhibit good initial skills with Cor exceeding 0.6 over most regions at 1-week lead, and the skills gradually decayed with lead time. By week 4, mid-to-high latitude predictability diminished substantially (Cor<0.2), while certain tropical regions maintained moderate skills (~0.5).

In-depth analysis of tropical regions revealed that the prediction skills were primarily modulated by the annual cycle and high-frequency components (3–10 days). The annual cycle component exhibited strongly positive correlation with predictability (the correlation coefficient was 0.84), whereas high-frequency activity exhibited a robust negative correlation (the correlation coefficient was −0.73), both exceeding the 99.9% significance level. This demonstrated that the enhanced S2S predictability skill in regions dominated by the annual cycle and reduced skill where high-frequency variability prevailed. Singular Value Decomposition (SVD) analysis indicated that the annual cycle components of tropical WPD were primarily linked to the annual cycle of solar radiation, while high-frequency activities were closely associated with tropical wave dynamics.