Prototyping Multi-model Ensemble Forecasts of Ground Magnetic Disturbances

In our dynamic space weather environment, there are many impacts of space weather on our current infrastructure. Notably, ground magnetic disturbances (GMD) that can interrupt and damage power grids. The ability to provide substantial space weather forecasts of GMDs is paramount. Ensemble forecasting provides a way to decrease uncertainty in forecasts and is commonplace in meteorology. While the use of ensembles have been increasing for space weather forecasting, multi-model ensembles for GMDs have yet to be implemented.

This study examines the efficacy of multi-model ensembles for GMD forecasting. Following the same validation process that led to the implementation of the Space Weather Modeling Framework (SWMF) at SWPC. Utilizing the open data set from Pulkkenin et al. 2013, results combined five models to produce a multi-model ensemble via a naive probabilistic classifier (NPC). Analysis of the ensemble was compared to the SWMF with three metrics and one skill score. The approach taken in the creation of the NPC was based on a set number of models crossing set thresholds in the same time frame. Overall the models’ included all have a predilection towards underprediction, and this is apparent in the lowering in the Heidke Skill Score and in the Bias. However, looking at specific events and magnetometer groupings improvements can be found. Strong improvements are seen for specific events or magnetometer groupings, such as events 4 and 8.

Being able to show feasibility of ensemble models in space weather forecasting is an important next step for the community as we explore new methodologies and applications.