We have analysed the statistical association between satellite anomalies and magnetic storms identified using our Dst-based storm definition (see How Dst relates to storms.) We found that there is an increased likelihood of satellite failiure in the six-day period following a storm onset.
Based on this association, and using our predictions of the Dst index for the present time, we forecast satellite risk for the next six days in the following way:
We have shown that this forecasting method has a certain level of skill at predicting satellite anomalies by looking at electrostatic discharge anomalies in a satellite anomaly database available from NOAA.
See the current satellite risk forecast.
The graph above shows the behaviour of the Dst index during a typical magnetic storm. The coloured bar at the right of the plot shows activity levels that we have defined. To see how we use these activity levels to make forecasts of satellite risk, we consider the forecasts made at five times a-e during the time interval shown above.
At time a (well before the start of the storm) the current activity is low and has been low for some time. Consequently, we say that the risk for the current hour (time a) is low and that the risk for the following six days is also low. Note that when we make the forecast at time a, all we can see is the trace up to time a and not afterwards, there is not yet any evidence for an imminent storm.
Around a day later, at time b, Dst has begun to suddenly drop, into the medium activity zone. All we know at this time is that Dst has begun to drop, which might mean that a storm is beginning, but might simply be a minor fluctuation. To reflect the uncertain nature of this time, we describe this time period as medium risk, and change our prediction for the following six days to medium risk.
By time c, Dst has dropped well into the high activity zone, and we know a fully fledged storm is in progress. We know that the period of highest risk to satellites is the six days from the onset of the storm (i.e. from time b until exactly 6 days after time b.) However, time b is in the past, and so we can't change the risk index we gave to that time period. So we leave the risk index from time b until just before time c as medium risk, and predict a risk index of high risk from time c until exactly 6 days after time b. This six day period is now absolutely fixed; whatever Dst now does the risk index for this time will remain high. But what about the time after this six-day block? For now, while Dst remains in the high activity zone, we will continue to predict that these time periods, which are almost 6 days in the future, will be high risk, but reserving the right to change our forecast as more data is available.
Time d is well into the recovery phase of the storm, and the Dst index has returned to the medium activity zone. However, we are still well within the six-day danger period for satellites which is associated with the storm. So, as explained above, we do not change our predictions for the six-day block beginning at time b, and thus time d is described as a high risk time period. Again, what about the time after this six-day block? Because we are in a period of medium activity, it is equally possible that the immediate future is going to consist of another storm, or a quiet period. We thus forecast medium risk.
At time e, the storm has ended and Dst is back in the low activity zone. This time is still within the six-day high risk time period which began at time b, and so will be described as high risk, but now that the evidence is that the storm will not immediately be followed by another, we predict low risk for the time after the six-day high risk period.
Remember that the graph above shows the behaviour of the actual Dst index. At the time we make real forecasts, we use an estimated value of Dst that has been predicted from an earlier recorded value. Thus the satellite risk forecast relies on the accuracy of our Dst predictions. Available is an comparison of the observed and predicted Dst values over the last month.