NASA sun data helps new model predict big solar flares

NASA/GODDARD SPACE FLIGHT CENTER

Using data from NASA’s Solar Dynamics Observatory, or SDO, scientists have developed a new model that successfully predicted seven of the Sun’s biggest flares from the last solar cycle, out of a set of nine. With more development, the model could be used to one day inform forecasts of these intense bursts of solar radiation.

IMAGE
IMAGE: AN X-CLASS SOLAR FLARE FLASHES ON THE EDGE OF THE SUN ON MARCH 7, 2012. THIS IMAGE WAS CAPTURED BY NASA’S SOLAR DYNAMICS OBSERVATORY AND SHOWS A TYPE OF LIGHT… view more CREDIT: NASA’S GODDARD SPACE FLIGHT CENTER/SDO

As it progresses through its natural 11-year cycle, the Sun transitions from periods of high to low activity, and back to high again. The scientists focused on X-class flares, the most powerful kind of these solar fireworks. Compared to smaller flares, big flares like these are relatively infrequent; in the last solar cycle, there were around 50. But they can have big impacts, from disrupting radio communications and power grid operations, to – at their most severe – endangering astronauts in the path of harsh solar radiation. Scientists who work on modeling flares hope that one day their efforts can help mitigate these effects.

Led by Kanya Kusano, the director of the Institute for Space-Earth Environmental Research at Japan’s Nagoya University, a team of scientists built their model on a kind of magnetic map: SDO’s observations of magnetic fields on the Sun’s surface. Their results were published in Science on July 30, 2020.

It’s well-understood that flares erupt from hot spots of magnetic activity on the solar surface, called active regions. (In visible light, they appear as sunspots, dark blotches that freckle the Sun.) The new model works by identifying key characteristics in an active region, characteristics the scientists theorized are necessary to setting off a massive flare.

The first is the initial trigger. Solar flares, especially X-class ones, unleash huge amounts of energy. Before an eruption, that energy is contained in twisting magnetic field lines that form unstable arches over the active region. According to the scientists, highly twisted rope-like lines are a precursor for the Sun’s biggest flares. With enough twisting, two neighboring arches can combine into one big, double-humped arch. This is an example of what’s known as magnetic reconnection, and the result is an unstable magnetic structure – a bit like a rounded “M” – that can trigger the release of a flood of energy, in the form of a flare.

Where the magnetic reconnection happens is important too, and one of the details the scientists built their model to calculate. Within an active region, there are boundaries where the magnetic field is positive on one side and negative on the other, just like a regular refrigerator magnet.

“It’s similar to an avalanche,” Kusano said. “Avalanches start with a small crack. If the crack is up high on a steep slope, a bigger crash is possible.” In this case, the crack that starts the cascade is magnetic reconnection. When reconnection happens near the boundary, there’s potential for a big flare. Far from the boundary, there’s less available energy, and a budding flare can fizzle out – although, Kusano pointed out, the Sun could still unleash a swift cloud of solar material, called a coronal mass ejection.

Kusano and his team looked at the seven active regions from the last solar cycle that produced the strongest flares on the Earth-facing side of the Sun (they also focused on flares from part of the Sun that is closest to Earth, where magnetic field observations are best). SDO’s observations of the active regions helped them locate the right magnetic boundaries, and calculate instabilities in the hot spots. In the end, their model predicted seven out of nine total flares, with three false positives. The two that the model didn’t account for, Kusano explained, were exceptions to the rest: Unlike the others, the active region they exploded from were much larger, and didn’t produce a coronal mass ejection along with the flare.

“Predictions are a main goal of NASA’s Living with a Star program and missions,” said Dean Pesnell, the SDO principal investigator at NASA’s Goddard Space Flight Center in Greenbelt, Maryland, who did not participate in the study. SDO was the first Living with a Star program mission. “Accurate precursors such as this that can anticipate significant solar flares show the progress we have made towards predicting these solar storms that can affect everyone.”

While it takes a lot more work and validation to get models to the point where they can make forecasts that spacecraft or power grid operators can act upon, the scientists have identified conditions they think are necessary for a major flare. Kusano said he is excited to have a promising first result.

“I am glad our new model can contribute to the effort,” he said.

###

From EurekAlert!

21 thoughts on “NASA sun data helps new model predict big solar flares

  1. They better hurry up because it looks like Solar Cycle 25 is ramping up… Reason… It started earlier like NASA predicted…

    When did Solar Cycle 25 start?
    We developed a new theory (A Formula For the Start of a New Sunspot Cycle) to calculate the start of a new sunspot cycle: the paper was published in Astrophysics and Space Science. Determining the start of a solar cycle is one of the most followed questions in astrophysics because it may be important to professionals like astronauts, astrophysicists, engineers responsible for protecting the power grid, etcetera.
    The latest NASA prediction panel considers April 2020 as likely to become the starting month of the new cycle. We disagree and point to October 2019 as a central point to calculate the start. Why? Since 1947 a radio telescope in Canada has been measuring solar flux. We found something peculiar: in most of the previous 6 cycle transitions, the lowest daily solar flux values were near 64. The new solar cycle started a few months before or after these clusters of minimum values. In October 2019 there was another cluster of measurements below 66. A preliminary conclusion was that Cycle 25 was going to start between August 2019 and January 2020.
    Co-author Jan Alvestad has a widely followed website Solar Terrestrial Activity Report and maintains high resolution sunspot counts based on images from the SDO NASA spacecraft. If you look (indirectly) at the Sun with telescopes, most days will be spotless near solar minimum, and those spots that can be observed are small and usually disappear quickly. However, there are plenty of tiny spots in high resolution images. For instance when other observers using traditional resolution telescopes see 1 sunspot at minimum, Jan Alvestad observes and documents 4-6 times more at the highest image resolution. This gives a new perspective on the 300 year old method of counting sunspots.
    Meanwhile we found more markers (under review) and their latest calculations point to November-December 2019, and especially December 2019 as the likely start of Solar Cycle 25.
    Shortly after we found that Solar Cycle 25 started in November or December 2019, we discovered something that at first seemed hard to believe. Using 365 days smoothing, 4 out of 5 of the data series available all had the solar minimum on the same day. The NOAA sunspot number, solar flux at 1 AU as well as both the STAR 1K and 2K high resolution sunspot numbers all had their lowest value on November 17, 2019. We sent a paper on this discovery for peer review knowing it would not be published before the official announcement of the start of Solar Cycle 25. Anyway, co-author Jan Alvestad added this important information to the STAR web site in June 2020. The pre-print was published on ResearchGate as the last in a trilogy of papers that could change how we determine when a new solar cycle begins.
    More can be found on the website of Jan Alvestad: http://solarresearch.info/

  2. While it takes a lot more work and validation to get models to the point where they can make forecasts that spacecraft or power grid operators can act upon, …

    Really?

    In 1989, a power blackout in the American northeast and Canada was triggered by a geomagnetic storm that overloaded one part of the power grid and caused a blackout to cascade through the system. Several satellites have been disrupted as high-energy particles associated with the solar wind flowed through sections of the satellites and damaged their sensitive electronics. In 1979, the Skylab space station prematurely re-entered Earth’s atmosphere due to a malfunction caused by increased solar activity, and consequently rained debris over the Indian Ocean and parts of western Australia. link

    Hmmm …

    Can’t you see a CME coming with enough time to shut down the grid? How much extra time would a successful forecast give us? Hasn’t the grid been considerably hardened in the last thirty years?

    Is the ability to forecast CMEs being oversold? I have no clue.

    • “Shut down the grid” … LOL…. you funny man.

      The intent is to engineer in enough over current and over voltage protection elements to not have the geomagnetic storm shut down the grid.

      The big CME’s can arrive here in 18 -24 hours. Maybe a super energetic one in 16 hours (2,500 km/sec). We know they are coming. We can see them lift away from the corona. The biggest thing that needs to happen is satellite operators be warned so they can take preplanned measures on their satellites to try to protect them.

  3. a new model that successfully predicted seven of the Sun’s biggest flares from the last solar cycle, out of a set of nine.

    “Predicting” past events is a lot easier than predicting future ones.

    NASA need to check the meaning of the word predict.

    • I think the article author misunderstood that the model was being trained by using past events. It’s vital that the model be tested for learning–that’s how good models are built in machine learning.

      The problem doesn’t seem to be the model, from what I know and have understood, they successfully trained the model with a pretty high confidence level. This should aid (not be an absolute, but aid) in helping to predict the probabilities of a solar flare in the future.

      But that doesn’t sound as sexy and attention getting as NASA can predict solar flares now! Go NASA! Team NASA! etc…you get my drift on that one. At least the author included the quote from the scientist about the model helping and not an absolute. Which is what a model is supposed to do..HELP, not be an absolute for massive decision making and burning money to feed a political policy furnace.

    • Models don’t know what the day is. If you feed the model historical data, it can then try to predict what will happen next. If what the model predicts matches what actually happened, then you have a successful prediction. Whether that prediction was the result of a good model or mere circumstance, can be determined by more testing.

    • I won’t pay for the article, so I can’t determine what is meant by predict.

      https://science.sciencemag.org/content/369/6503/587

      Anyone with access please clarify.

      I will also point out that the authors are esteemed Japanese scientists, and that there is a possibility that a nuance in Japanese may have not been expressed properly in English.

      There are other nuances that confuse me in what I can read from NASA and SCIENCE. There were 50 of the events they are predicting. They restricted themselves to specific regions, reducing that number. They made 10 predictions watching 7 hotspots. 3 predictions were false. There were 9 qualifying flares. 2 were from hotspots that did not meet their criteria for watching.

      With such a long data time, it is possible that criteria were predetermined prior to analysis, but long after actual events. In that case, I find it hard to believe that they didn’t know about the two their criteria missed before they started.

      And, if it is a ‘prediction’ from past data, it would be good to know the feasible warning time possible if applied to live data. Mechanics of data collection. Analysis time to find criteria. Simulation time to predict event (if necessary). Each of these things could delay and prevent real prediction.

  4. #greg

    You are expressing the same thought I had reading the headline.

    A successful model ist validated by the ability to correctly predict future event’s!! Not the past because you can tweak any model the match „some“ past events (7 out of nine). Yes, it’s a good start but nothing more and it’s not a validation!

    • Sometime (but not very often) if you get lucky it predicts future far more accurately than you ever hoped for. More than 10 years before SC24 hit its maximum annual value someone out there predicted it’s value within a decimal point, as you can see here
      http://www.vukcevic.co.uk/SSN.htm
      or if so inclined check it for yourself.
      The equation has been tried for lottery, stock market, horse racing, even next month’s daily temperatures, but failed every time, hence it must have been an ‘incredible coincidence’. 🙂

    • I’m so disappointed. I thought they must have discovered time-travel so they could go back in time to predict what was then future events, and then return to the present to report their findings so as to not disrupt the timeline. As far as I can see, it is not really a model anyway, but more of an early warning system to detect a precursor to a solar flare.

    • When you can predict nine of the last seven events you really have something!

      Getting a model right is an iterative process. First, you are always using past events to create your model so yes, it is going to match it otherwise it is dead out of the box. The next step is to wait for the event to happen and see if your model predicted it. Even if it did, you still want to review it and see if you can add/change something to the model to make it more accurate in the future.

      Take a look at the weather prediction models. They do a pretty good job predicting three days out. Go to five days and not so much. Are they valid models? Yeah but with reservations.

  5. It is unclear from the article how much lead time the prediction gives. Is it enough to make a difference or not?

Comments are closed.