How we predict Dst

Why do we need to predict Dst?

We have developed a method to forecast the short-term space weather risk to satellites based upon the current value of Dst. However, Dst measurements are not instantly available, because preparing it is a complex process. Several versions of Dst are released, the earliest being quick-look Dst which is released between 12 and 36 hours after the measurements are made.

To predict the current value of Dst we use the most recent quick-look Dst value and real-time solar wind data measured by the ACE satellite, using a method based upon that of O'Brien and McPherron.

Applying the method

The practical application of the method involves several stages, augmented by test criteria for accuracy of the Dst prediction, which in turn determine the timing of the process.

1. Retrieval of solar wind data - ACE satellite data will be used as it is available with only a few minutes lag of real time, and it is freely available.
2. Propagation of the solar wind from the ACE satellite location to the Earth using a method of Lester et al.
3. Calculation of hourly averages of the solar wind data for use with O'Brien and McPherron's method.
4. Retrieval of the most recent Dst data available from the World Data Centre for Geomagnetism, Kyoto, Japan.
5. Calculate Dst^* by applying the pressure correction to the Dst values from the WDC.
6. Using the most recent Dst^* value as the initial condition, use the Burton equation with hourly VBs from the propagated hourly averaged ACE data to predict values of Dst^* for every hour up to the present hour in UT time for a nowcast and 1 hour forward for a forecast. Data gaps in ACE are accounted for by interpolation, where possible, or by extending the time step in the Burton equation.
7. Use the ACE data to inverse pressure correct the predicted Dst^* values to obtain predictions of Dst. Data gaps in ACE are accounted for where possible by interpolation, or by use of the previous ACE values.

The timing of the process is determined automatically with the aim of performing the nowcast as soon as is realistically possible. Attempts at a nowcast, subject to the following tests, will be made half an hour after the last successful nowcast. Nowcasting is defined as predicting the next hour on from the previous nowcast irrespective of the actual time. Any break down in the program will result in the program being reset and the nowcast will be set to the present hour.

The process described above is augmented by several approximations, restrictions and tests must be made due to the nature of the data available.

The data set of solar wind properties from the ACE satellite contains data gaps ranging from several hours of missing data to the occasional minute of bad data. The solar wind parameters necessary are the solar wind speed v, the density of the solar wind n and the z component of the interplanetary magnetic field. Solar wind speed and density are recorded by the SWEPAM instrument, B_z by the MAG instrument, both instruments give minute values but data gaps and the final minute given by SEC are not always concurrent.

The satellite risk assessment is based on the nowcast of Dst. The aim being to make this nowcast as soon as possible, while maintaining an acceptable level of accuracy. Thus several criteria are applied to the solar wind parameters from SEC before hourly averages are calculated.

• The SEC data is only included if the data is considered "good" by the standards set by SEC. Further information about the bad data may allow it to be included in our study at a future time.
• The hour to be nowcasted must have a complete hour of solar wind data available (allowing for data gaps), i.e. the time of the final value of ACE solar wind data when propagated to the earth must be later than the end of the hour to be nowcasted.
• To calculate any hourly average there must be solar wind data present for at least 20 minutes of the hour, else a data gap is recorded. This is an arbitrary condition that could potentially be relaxed.

The first and third conditions are regarded as non-critical in that a data gap in the ACE data will be recorded for the nowcast time and a nowcast of Dst will be made by using a longer time step in the Burton equation.

The second criterion is regarded as the critical condition, such that delaying the nowcast could possibly make a better more accurate nowcast. As such from this criterion a delay time is calculated after which an attempt at calculating the nowcast will be re-attempted. The delay time is given by the end of the UT hour wishing to be nowcasted minus the time of the final value of solar wind data propagated to the earth.

The Results

Plot showing the values of Dst predicted using this method

The point within the vertical dashed lines is the prediction of Dst for the latest hour we can make a prediction for - the "nowcast." It is this value which is stored as our prediction for any future analysis. Note the predictions up to this nowcast from the final quick-look Dst value are not fixed but may change as quick-look Dst is updated or as more solar wind data is obtained allowing interpolation across data gaps. However the nowcast made for each hour is stored as is, and will not be updated even if subsequently an improved prediction for that hour (now lagging real time) could be made.

The plot highlights the 4 data conditions

• Blue - Dst is estimated using a single hour time-step with ACE data present for the hour being predicted
• Orange - ACE data for the hour is not present in the data record and is estimate via interpolation across the data gap
• Yellow - ACE data for the previous hour is not present and has been interpolated
• Red - Both this and the previous hours ACE data are not present in the original data and are interpolated

How we've done in the last week
How we've done in the last month