Satellites like DSCOVR are used to predict solar storms.

How does that work? If the satellite is orbiting L1 and "watches" the sun, how can the signal be back on earth before a storm arrives?

How are these storms predicted exactly? Are the models like for weather on earth?

Can these storms be predicted several days before, like with weather on earth?


3 Answers 3


You may be asking about a couple of different things here.

1 Events at the surface of the sun

Events at the Sun itself may or may not head in the direction of the Earth, but this is step one of a prediction system. I believe there are two approaches:

a) look at what happened about a month ago and predict that it will be still there when that part of the Sun rotates back into view. I think this has been tried for a while as it is conceptually straightforward.

b) try and forecast an event (e.g. flare or CME) from the appearance of other measurable features on the Sun. I believe this is pretty involved in modelling terms and has not yet come of age (anyone know better?)

2 Events close to the Earth

A geomagnetic storm is the term given to phenomena measurable at the Earth but caused by previous events on the Sun. Whilst a flare (e.g. X rays) might travel quickly to Earth the accompanying release of particulate matter can take of the order of three days to transit.

a) Having a clear view of the sun in several wavebands does therefore allow quite an advanced warning but does not prove that the event seen at the sun is heading towards the Earth (cue more modelling work) or will connect magnitically with the Earth's field when it does (local measurements needed).

b) Having instruments at L1 for sampling of the local environment there may well say much more about the disturbances in the local region of space but unfortunately at not much advanced a warning (a few hours).


Solar flares travel at about 2000 km/s. Radio travels at 300,000 km/s, so a spacecraft at L1 (1.5 million km out) like DISCOVR gives about 12 minutes of warning when the solar flare reaches it.

Spacecraft like SOHO observe the sun, and these observations are used by e.g. NOAA to try and predict flares:

Current methods of flare prediction are problematic, and there is no certain indication that an active region on the Sun will produce a flare. However, many properties of sunspots and active regions correlate with flaring. For example, magnetically complex regions (based on line-of-sight magnetic field) called delta spots produce the largest flares. A simple scheme of sunspot classification due to McIntosh, or related to fractal complexity.[57] is commonly used as a starting point for flare prediction.[58] Predictions are usually stated in terms of probabilities for occurrence of flares above M or X GOES class within 24 or 48 hours.

  • $\begingroup$ Solar flares are localized intensifications in EM radiation, usually in the UV and x-ray bands. So they do not travel at 2000 km/s but at c. I think you meant really fast coronal mass ejections or CMEs (since most CMEs only propagate at ~300-400 km/s) there. $\endgroup$ Commented Sep 2, 2021 at 13:26

Satellites like DSCOVR are used to predict solar storms.

No they are not. Spacecraft cannot predict solar storms. They can inform a user that something is coming toward Earth, but they cannot predict the onset of solar phenomena. That is, spacecraft are capable of providing a warning but they cannot predict anything (I will distinguish that users interpreting data also suffer the same fate with solar phenomena later).

Remember that the fastest information gets to spacecraft near the first Lagrange point, L1, at the speed of light. So things like solar flares cannot be predicted as they are just localized intensifications in electromagnetic (EM) radiation (usually dominated in UV and x-ray bands). That is, by the time we "see" the flare it is already too late.

Why you ask? Well the spacecraft makes a measurement, then the electronics process it. Even if we have a real-time telemetry stream, there will still be several micro- to milliseconds of delay between the observation and transmission. The transmission moves at the same speed as the solar flare emissions, but is now at least several kilometers behind the leading edge of the solar flare emissions. Then a ground station receives the telemetry packets and decodes them, then re-packetizes them and sends them along to the user. The user receives said packetized files and then opens the files, calibrates the data, and plots the results. All of these steps once the spacecraft transmission hits the ground station can take many seconds to several minutes or more.

So for the EM radiation from a flare, we will never be able to warn the user.

However, there are often phenomena like solar energetic particles (SEPs) that result from solar flare (and related phenomena like coronal mass ejections (CMEs)). The electron SEPs usually arive shortly after the EM radiation from the flare. Shortly here depends on the energy of the electrons and the delay in release time of the electrons (i.e., the processes that accelerate electrons do not necessarily occur simultaneously with those that emit the EM radiation). Even so, the highest energy electrons should get to Earth in just over eight minutes, i.e., slightly longer than the EM radiation. This is moot since DSCOVR does not have high energy electron telescopes, but we will ignore this for the moment. Regardless, we can see that electrons are not good for a warning system either.

So how about energetic ions? Well here we may be better suited as these can take significantly longer to reach Earth than the electrons or EM radiation (i.e., minutes). The most energetic of these (i.e., ~1 GeV protons move at ~88% of the speed of light) will get to Earth very quickly, but the lower energy particles take much longer. So if we see enhanced, localized EM radiation in the UV and x-ray bands in addition to energetic electrons (i.e., several 10s of keV to 100s of keV), we can expect energetic ions to arrive shortly after and potentially warn customers. However, this is also unlikely due to the delay between spacecraft measurement and user plotting that I mentioned above.

So is there any hope? Perhaps the most beneficial and only feasible avenue (at least for now) would be to use a spacecraft with a coronagraph like SOHO with a high transfer rate real-time telemetry stream and automated ground transfer and plotting schemes (i.e., minimize slow human steps as much as possible). However, even this has the uncertainty of not knowing whether an observed CME is directed toward or away from Earth. In the 2D projection these imagers generate, such a CME looks like an expanding ring, thus the name halo CME. For strong events that are Earth-directed, the images are often accompanied by a phenomena called snow. This is just energetic particles hitting the CCDs in the camera causing spots and streaks to occur in the image. Those generally do not happen when the halo CME is directed away from Earth (but not always).

So what are we left with? Suppose the coronagraph method works and suppose we know the CME is directed toward Earth, then what? Do we warn our customers to shutdown spacecraft or go into safemodes to protect internal systems immediately? The fastest CMEs move at over 2000 km/s, which means they get to Earth in just under 20 hours. They can propagate from L1 to Earth in around ~12 minutes, so we cannot really wait until the L1 monitors see them if it's a fast one. How do we know it's a fast CME? We don't really, without additional info. The 2D projections in coronagraph images are difficult to interpret and can give unconstrained estimates without additional observations.

So can we predict these phenomena ahead of time? Nope, not to any level of statistical certainty. The best we can do is observe a large and strong sunspot and say something like, "That thing is going to erupt..." Okay, we can do a little better but that's another several pages of nuances and caveats. But the biggest and strongest sunspots don't always generate the strongest flares or CMEs. Sometimes they generate lots of smaller events but no big events (at least none on the sides of the Sun we can observe) while others generate several small and medium events with a few large ones. It's really not possible to predict exactly what the solar active region will do with any level of confidence.

If you are familiar with the difficulties involved in weather forecasting, you may appreciate the following. So you have probably heard of scattered thunderstorms, which are the bane of a weather forecasters' existence. The reason being that they are really unpredictable beasts. They can arise from or be killed by micro- and mesoscale fluctuations. Even with thousands upon thousands of measurement locations simultaneously over a small region of land, we still cannot predict these things other than to say that conditions are good or bad for them to generate.

In space we have less than a dozen spacecraft dedicated to observing the sun and solar wind directly. The volume of space to monitor is orders and orders of magnitude larger than one county in a USA state, right, yet we have only three real-time monitors (i.e., SOHO, ACE, and DSCOVR) and several more in situ spacecraft (e.g., Wind, STEREO, Parker Solar Probe, Solar Orbiter, etc.).

To put this in perspective, weather forecasts in the 1950s were more accurate than our best efforts with space weather now for two primary reasons: we need 1000s more observation points to even compete and solar weather includes magnetic fields adding at minimum three more equations to solve for modeling (usually it's six or more).

So no, we cannot predict space weather phenomena with any reliable accuracy.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.