How Can Turbulence Improve Space Weather Forecasts?
By Cara Waters (Queen Mary University of London)
The auroral electrojet (AE) index is one of the key indicators of high latitude geomagnetic activity. Forecasting it from upstream solar wind conditions remains challenging due to nonlinear coupling and multiscale variability. To combat this, we test whether incorporating solar wind turbulence improves short timescale AE forecasts beyond models based on solely the mean solar wind and interplanetary magnetic field. We compare two gradient boosted decision tree (XGBoost) models using near-Earth solar wind observations from OMNI. The baseline model uses standard mean field, density and velocity parameters, while the turbulence-aware model adds measures of fluctuation amplitude, intermittency, energy partition, compressibility, and Alfvénic structure. Both models achieve a peak performance at short lead times, giving correlations above 0.8 at 60 minutes. However, the turbulence-aware model maintains forecasting skill for longer lead times. The turbulence-aware model also provides consistent improvements over both the baseline model and persistence. Critically, this model improves forecast robustness for high-impact events. Using cost-loss analysis, the baseline model provides decreasing economic value with increasing AE threshold. In contrast, the turbulence-aware model maintains an approximately constant threshold for positive economic value, indicating stable economic usefulness even for extreme AE conditions. Using interpretable machine learning techniques, we can show that the most important turbulence parameters for this at short timescales are fluctuations in Bz and the skew of these fluctuations, and at longer timescales it is the properties of the turbulence (cross helicity, compressibility, and residual energy). This demonstrates that turbulence provides complementary, scale-dependent information beyond mean solar wind parameters, improving both forecast performance and decision-relevant value for operational space weather applications.
See publication for more details:
Cara L. Waters, Christopher H. K. Chen, Mathew J. Owens (2026). Beyond Mean Solar Wind Conditions: Turbulence-Aware Forecasting of the AE Index. Space Weather, 24(7). https://doi.org/10.1029/2026SW005094
Potential economic value 𝑉 against cost/loss ratio 𝑟 for the base model in (a) and turbulence model in (b), for a range of thresholds of auroral electrojet (AE) between 800 and 1,200 nT. (c) The 𝑥-intercept of each of the curves plotted against the threshold AE index for the turbulence model (red circles) and the base model (blue squares).