Evaluating Auroral Forecasts Against Satellite Observations

The aurora is a readily visible phenomenon of interest to many members of the public. However, the aurora and associated phenomena can also significantly impact communications, ground‐based infrastructure, and high‐altitude radiation exposure. Forecasting the location of the auroral oval is therefore a key component of space weather forecast operations. A version of the OVATION‐Prime 2013 auroral precipitation model (Newell et al., 2014, https://doi.org/10.1002/2014sw001056) was used by the UK Met Office Space Weather Operations Centre (MOSWOC). The operational implementation of the OVATION‐Prime 2013 model at the UK Met Office delivered a 30‐min forecast of the location of the auroral oval and the probability of observing the aurora. Using weather forecast evaluation techniques, we evaluate the ability of the OVATION‐Prime 2013 model forecasts to predict the location and probability of the aurora occurring by comparing the forecasts with auroral boundaries determined from data from the IMAGE satellite between 2000 and 2002. Our analysis shows that the operational model performs well at predicting the location of the auroral oval, with a relative operating characteristic (ROC) score of 0.82. The model performance is reduced in the dayside local time sectors (ROC score = 0.59) and during periods of higher geomagnetic activity (ROC score of 0.55 for Kp = 8). As a probabilistic forecast, OVATION‐Prime 2013 tends to underpredict the occurrence of aurora by a factor of 1.1–6, while probabilities of over 90% are overpredicted. Plain Language Summary Enhanced auroral activity at Earth can cause disruption to long‐range radio communications and ground‐induced currents making forecasting the location of the auroral oval and probability of the aurora occurring of interest to many sectors such as aviation, energy, and defense. The UK Met Office uses a version of the OVATION‐Prime 2013 auroral forecast model to deliver a 30‐min forecast of the location and probability of observing the aurora. In this study, we evaluate the performance of the auroral forecasts against satellite observations of the aurora, captured by the IMAGE satellite between 2000 and 2002. Our analysis shows that the auroral forecast model performs well at predicting the location of the auroral oval, under nominal space weather conditions, but the probabilities of aurora occurring forecast by the model tend to be underpredicted, in other words, the aurora occurs more frequently than the forecast mod