Analog ensemble forecasts of solar wind parameters: Quantification of the predictability and time-domain spectral performance

Abstract

Forecasting multiscale properties of the solar wind is one of the important aspects of space weather prediction as mesoscales, larger than one minute, can affect the magnetosphere. Amongst forecasting techniques, the Analog Ensemble (AnEn) method allows the forecast of a quantity from its past behavior, is easy and quick to implement, and results in an ensemble of time series. A comparison of optimal AnEn forecasts of \textit{Wind} spacecraft observations of near-Earth solar wind properties with the persistence and climatology baselines allows a quantification of the predictability of the magnetic and velocity components and magnitude. The AnEn predictions were found to be as accurate as persistence for short-term forecasts and climatology for long-term ones, and performed better than both baselines for more than 60\% of the samples for a particular lead time. Furthermore, using an AnEn instead of the baselines enables prediction of the full spectrum of solar wind fluctuations. However, using the standard averaging method to generate a unique forecast from the AnEn ensemble results in a loss of power in the small-scale fluctuations. To prevent this loss, a new spectral reduction method is proposed and compared to the standard averaging method as well as the synodic recurrence baseline. The AnEn spectral-reduced forecast is shown to be more time-accurate than the synodic baseline and more frequency-accurate than the mean-reduced forecasts. Such a reduced forecast is then confirmed to be useful as a comparative baseline in performance diagnostics of space weather models.