CME Tracker help
What this is
CME Tracker is a live view of space weather between the Sun and Earth. It shows where coronal mass ejections are, where they are headed, and whether and when they will reach us.
A coronal mass ejection (CME) is a cloud of magnetized plasma blasted off the Sun. When one aimed at Earth arrives, typically 1 to 3 days later, it can compress Earth's magnetic field, light up auroras, and disrupt satellites, radio and power grids. The tracker lists every CME NASA has catalogued in the last 90 days, draws each one as an expanding cone on a map of the inner solar system, and estimates its arrival. New to CMEs? Start with What is a CME?
Each topic below is short. Where there is more to say, a More detail line opens the full explanation.
Quick start
The tracker answers one question first: is anything from the Sun headed for Earth, and when does it arrive? Five steps get you there.
- Read the list on the left. Each row of the CME Catalog is one eruption: its launch date, its speed, a one-word verdict for Earth, and the expected arrival if it has one. A red dot and the word Directed mean it is aimed at us.
- Click a row. The map jumps to that CME in mid-flight and plays forward. The Sun is in the middle, Earth is the blue dot on its orbit, and the CME is the colored wedge moving outward. Color is speed.
- Read the answer on the right. The Selected CME tab opens with the verdict, then the expected arrival with a countdown and its 80% window, then the likely storm strength. The details follow underneath.
- Move through time. space plays and pauses, and you can drag the timeline under the map. The buttons either side of Play jump to the next launch, arrival or flare. LIVE (or N) returns to now.
- Look around. The other tabs on the right are Earth (the globe, the magnetosphere, Kp, satellite drag and radiation), Sun (SDO images, X-ray flux and the coronagraph) and Solar wind (what is being measured at L1 right now). Above the map, Layers adds overlays and the view buttons switch to 3D, Magnetosphere and Atmosphere.
Nothing headed for Earth is the normal state. To see what a big storm looks like, open one from Famous CMEs and it replays on the same map.
On your first visit the tracker shows three short pointers of its own. For a guided version, the six-module course puts the live map on the page with one thing to do per module, about 40 minutes end to end. Teachers: there are ready lesson plans.
The screen
The tracker has five areas. The rest of this page takes them in this order.

The tracker on a desktop screen.
- Across the top: the header readouts, the live numbers for the Sun and the solar wind. Most of them open a popup.
- On the left: the CME catalog, one row per eruption, and under it the list of recent flares.
- In the middle: the map. The toolbar above it holds the time zone toggle, the picture and video buttons, the Layers menu and the four views.
- Under the map: the timeline and its playback controls.
- On the right: five tabs. Selected CME, Earth, Sun, Solar wind, and About, which says what the site is and where its data comes from.
The small tabs at the left and right edges of the map hide either side panel to give the map more room.
Header readouts
The strip across the top shows the latest values: the daily sunspot number and F10.7 radio flux, then solar wind speed, density, Bz, Kp and Dst. A status line at the right says how fresh the data is. Click a readout, or focus it and press Enter, to open a popup that explains the number. Esc or a click elsewhere closes it.
Sunspots
Opens the whole surface of the Sun unrolled into a strip, with every numbered sunspot region of the last rotation on it. The lit middle is the side facing Earth today. The dark ends are the far side as it was last seen.
More detail
Click Sunspots to unroll the Sun: a strip of the whole surface by longitude with every numbered sunspot region of the last rotation on it. The lit middle half is the hemisphere facing Earth now, photographed today by SDO/HMI. The dark outer quarters are the far side as last photographed when it faced us, up to four weeks ago, so regions there may have grown or decayed, and a group that emerged on the far side is unknown until it rotates in (the blue bar marks the stretch STEREO-A can see). Hover a region for its size, magnetic class and when it rotates into or out of view. The same strip opens from the Unroll button on the Sun card.
Solar wind speed
The speed sets the warning time. The popup draws the path from the Sun to Earth with the wind now on its way, and marks where the wind arriving on each coming day is right now.
The header readout during a moderate 535 km/s stream. Click it to see what's in flight.
The corridor as a calendar: this parcel is seven hours into its 3.2-day trip, and the ticks show where the wind that arrives on each coming day is right now.
More detail
Solar-wind speed matters for one reason: it sets the warning time. So the popup draws the Sun→Earth corridor with the wind currently en route, as a clearly-labeled time-lapse (at true speed the crossing takes days): a bright parcel departs the Sun and crosses to Earth, its label counting the trip as it goes ("in flight 1.0 d · Earth in 2.2 d"). That's a 3 to 4 day journey at typical speeds, which is exactly the lead time this site's forecasts live on. Calendar ticks under the corridor turn it into a forecast timeline: each tick marks where the wind arriving on that day is right now. Monday's weather is already two-thirds of the way here, Wednesday's is just leaving the Sun.
The lapse factor is stated on-canvas where it applies, and the note flags the working assumption honestly: the conveyor holds today's speed for the whole trip, and real streams genuinely vary, which is exactly why the arrival days shift when the wind speeds up or slows down.
Density
Protons per cubic centimeter. The popup draws a one-centimeter cube holding that many protons. When a CME's leading edge arrives the count jumps, which is one of the first signs of impact.
7.6 protons per cubic centimeter. Click to see all eight of them.
7.6 p/cm³: eight protons in a sugar-cube of space. The same cube of the air you're breathing holds ~25 quintillion molecules.
More detail
Density is defined as protons per cubic centimeter, so this popup isn't a metaphor: it draws a 1 cm cube of space holding exactly the current reading's worth of protons, drifting through with the wind. The count follows the live number: when a CME's compressed sheath arrives and density jams from ~5 to 20 to 50+, you watch the cube crowd up, one of the first visible signs of impact.
The fine print keeps it honest: the dots are magnified ~10¹²× (a real proton at this scale would be invisible), the reading is an average so any actual cube fluctuates moment to moment, and every proton travels with an undrawn electron that keeps the wind electrically neutral.
Bz
The north or south tilt of the magnetic field carried by the wind, and the number that decides whether a storm happens. When Bz points south the wind's field connects with Earth's and energy gets in. When it points north, the same wind passes by. The readout turns red when it swings far enough south to drive a storm.
The readout turns red when Bz swings south enough to drive storming. Click for the gate.
A strongly southward spell (Bz −12.4 nT): the gate meter pegged at "wide open · G3+ if sustained," aurora caps glowing.
More detail
Bz gets the most consequential visualization because it is the most consequential number: it decides whether the wind's energy gets in at all. The popup animates dayside magnetic reconnection from the live field. Earth's dayside field lines always point north. That's the fixed door frame. The incoming field dashes are tilted to the measured Bz/|B|: pointing north, they can't reconnect, and you watch them slide around the magnetopause and past Earth. Gate closed, same wind, no storm. Pointing south, they splice with Earth's field at the nose (the flashes), and the opened lines peel poleward, funneling energy into the polar cusps and brightening the aurora caps.
The CLOSED→WIDE-OPEN meter is driven by the physical coupling term v·Bs, the same quantity the tracker's alert system watches at L1, mapped to the calibrated storm tiers. Its honesty notes ride along in the popup itself: a 2-D cartoon of real physics with compressed scales, reconnection is continuous (the snaps pace it, they don't count anything), and the reading is measured at L1, so the scene you're watching reaches Earth ~30 to 60 minutes later.
Kp
Global geomagnetic activity on a scale of 0 to 9, with a nine-segment meter beside the number. A storm starts at 5. The popup charts the past week of measured Kp, NOAA's three-day forecast, and markers where the tracker expects CMEs to arrive.
The meter during a G3 storm: seven segments lit, deep into the amber zone.
The past week of measured Kp, NOAA's forecast for the next three days, and arrival markers from the tracker's model for the CMEs inbound on October 8, 2026.
More detail
The header's Kp readout includes a mini-meter: nine segments spanning the full 0 to 9 scale, lit to the current value and colored along the quiet-green → storm-red ramp, so you can see at a glance where the current number sits in the range (storms begin at segment 5, where the ambers start).
Click the meter (or press Enter on it) and a panel opens charting the past week of measured 3-hour Kp alongside NOAA's official 3-day forecast. Solid bars are measurements, dimmed bars are NOAA's preliminary estimates, and hollow outlined bars are the forecast. Model output never masquerades as measurement. Amber bars show where this site's drag model expects inbound CMEs to arrive: the solid part is the estimated Kp and the faint dashed extension above it is the plausible upper bound, the same range the impact card shows, so you can see whether NOAA's forecast bump lines up with a CME we're tracking. Arrivals within eight hours of each other merge into one bar labelled with the count, because at the model's ±12 h they are one event, and a pile-up is what makes storms big.
Dst
How deep a storm is: how much the ring of charged particles circling Earth has weakened the magnetic field at the equator. More negative is deeper. The popup plots the past week against the storm thresholds and against famous storms.
More detail
Dst is the storm's depth rather than its activity: how much the ring current, the belt of charged particles circling Earth, has weakened the field at the equator. The popup draws that torus at its live strength and plots the past week of hourly Dst against the storm thresholds (−50 moderate, −100 intense, −250 super), with the phases labelled: the sudden commencement, the main-phase plunge, the slow recovery. A ladder alongside places today's value against the Gannon storm (−412 nT), Québec 1989 (−589), Carrington (estimated near −900) and the deepest hour of this year, so a number that means little on its own reads as a distance from history.
CME catalog and flares (left panel)
The catalog lists every CME from NASA's DONKI database over the last 90 days. The count beside the title shows how many are visible out of the total.
Reading a row
The catalog is a table with four columns: the launch date, the speed in km/s, the verdict for Earth, and the estimated arrival when there is one. Hover a row for the launch time, the speed class, the direction in solar coordinates (e.g. N12W34) and the cone half-angle. The verdict is one word with a coloured dot:
| ● Directed | The cone covers Earth. Expect an arrival. The arrival time is brighter while the CME is in flight. |
| ● Glancing | Earth sits near the cone's edge. A flank hit is possible but weaker. |
| ● Misses | Aimed away from Earth; no impact expected. |
| ○ Awaiting | Direction not triangulated yet. Too soon to call. It becomes No direction after about 4 days, when it's clear no analysis is coming. These never affect the Earth globe. |
Sorting. Click a column heading to sort by it, and click again to reverse. The list starts with the newest launch first. Sorting by arrival puts the CMEs still on their way at the top.
Click a row to select it (the map jumps to it and animates). Use ↑/↓ to move through the list.
Filters
- Speed class. The five chips at the top hide or show a class in both the list and the map. Hiding the slow S ones is the usual first move, since they rarely cause storms.
- Earth-directed only keeps just the CMEs headed for Earth, direct and glancing.
- Hide undetermined direction removes the CMEs whose direction NASA has not measured, the Awaiting and No direction rows.
The filters combine, and your choices are remembered.
Recent flares
- A flare tagged → CME launched a coronal mass ejection. Click it to jump to and select that CME. The matching burst pulses on the Sun image.
- Flares without the tag produced radiation but no catalogued ejection. Every row is still clickable: it jumps the playhead to the flare so you can watch it fire on the map and the Sun panel.
- Small
⌁II/⌁III/⌁IVchips on a row are the radio bursts logged with that flare (see Radio bursts under Layers). A Type II chip means a shock left the Sun, the earliest sign a CME launched; click the chip to scrub to the burst. - By default the list shows all flares. The toggle under the header flips to → launched a CME to show only the ones that produced a coronal mass ejection (the events relevant to tracking).
The map and its layers
The map looks straight down on the inner solar system.
- The Sun is at the center. Earth is the blue dot on its orbit, placed for whatever moment the clock shows.
- Each CME is a cone expanding outward in the direction NASA measured, as wide as NASA measured. A cone that covers Earth's dot will reach it. One pointing elsewhere passes by.
- Color is speed, cool for slow and hot for fast:
- A CME whose direction has not been measured yet is drawn as a faint full ring, not a cone.
- When an Earth-directed CME reaches Earth, the dot flashes.
Moving around
- Scroll or pinch to zoom, drag to pan. The circular arrow at the right end of the toolbar resets the view.
- Hover a cone for its speed, direction and arrival. Click it to select it.
The four views
The buttons beside the Layers menu change what the middle of the screen shows. Top-down is the flat map described here. 3D shows the same scene in three dimensions (see The 3D view). Magnetosphere and Atmosphere zoom in on Earth (see the Earth tab). Your choice is remembered.
Layers
The Layers menu in the toolbar switches overlays on and off. Radiation belts, Solar flares, Radio bursts, Labels and Cinematic view start on. The rest start off. Your choices are remembered.
The menu opens with three presets: Monitoring (the daily check: slow CMEs hidden, spacecraft on, pinned to live; key 1), Clean (just the map, both panels hidden; key 2) and Sensory overload (everything on; key 3). A preset is a starting point. Every switch can still be changed afterwards.
- Fluid model. Paints the solar wind across the whole map as a moving fluid, colored by speed or by density. It is computed in your browser from the wind measured at L1 and every CME on the map, so you can see fast wind catching slow wind and each CME's compressed front and thin wake.

Fluid model, speed. Fast CME material in green and yellow against the slow blue background wind measured at L1 over the last rotation.

Fluid model, density. The bright arcs are compressed sheaths the CMEs pile up ahead of themselves; the darker wake behind each is rarefied.
- Magnetic wake. The magnetic field measured at L1, drawn where that plasma is now. A ribbon trails outward from Earth: red where the field pointed south, which is what drives storms, blue where it pointed north, and faint where it was weak. Each CME that reached Earth leaves a wide striped fan. In the 3D view the same readings also stand on the plane as a curtain, below it for south and above it for north.

Magnetic wake during the May 2024 storm, from above. The deep red band is the southward field that drove it.
More detail
L1 measures the solar wind's magnetic field as it goes by. Each reading belongs to a parcel of plasma that keeps moving outward at the speed it was measured at, so a reading taken a day ago now sits about a day's travel beyond Earth. The ribbon strings those readings together for the last four days.
Nothing is drawn ahead of Earth, on purpose. A CME's field direction cannot be known until it reaches L1, 30 to 60 minutes before Earth, which is why storm strength is so hard to forecast. The ribbon starts at Earth because that is where knowing starts.
Two limits. L1 samples one line through each cloud. The fan is solid along Earth's line and fades toward its flanks to show that the rest is assumed. And a passage is matched to its CME by the forecast arrival time, moved to the nearest jump in measured wind speed, so CMEs that arrive close together can share one fan. Between storms the ribbon is a thin thread, because the quiet wind's field is weak.
- Streams. Fast wind from coronal holes, drawn as the teal spiral bands it forms. A stream is drawn once the wind at L1 has actually risen, and a dashed band marks where the same stream should return 27 days later. Live view only.

Streams, live view. Fast wind from a coronal hole is a spiral, not a straight jet, because the Sun rotates while the wind flows outward.
- Parker spiral. The spiral shape the Sun's magnetic field is pulled into by the wind. It is a faint backdrop for the other magnetic layers.

Parker spiral. Subtle on purpose; it's a backdrop for the other layers.
- Connectivity. The one magnetic field line that links the Sun to Earth, and the zone on the Sun around its base. An eruption inside that zone can send radiation to Earth within minutes, even if its CME misses.

Connectivity with a CME selected: the footpoint label, the well-connected fan, and the source flagged as connected.
- Spacecraft. Where the observatories this site reads from actually are: the fleet at L1, GOES and SDO at Earth, and STEREO-A, Parker Solar Probe and Solar Orbiter in their own orbits. Hover or tap one for what it does.

Spacecraft around Earth. Zoom in and the L1 cluster separates from Earth; STEREO-A's dotted sightline is its view of the Sun.
- Radiation belts. Draws the Van Allen belts around Earth in the Magnetosphere view. See Radiation belts under the Earth tab.
- Solar flares. Each flare flashes at the Sun when the timeline reaches its peak, colored by class (X red, M orange, C blue). A dashed ring means the flare launched a CME.

An X-class flare at its peak, with the dashed ring that means it launched a CME.
- Radio bursts. Radio sweeps logged by NOAA's ground stations, drawn at the Sun in teal. A Type II burst means a shock has left the Sun and is usually the earliest sign of a CME. When its source is known, the map draws a dashed provisional cone until NASA's measured one replaces it.

Radio bursts, zoomed in on the Sun at 08:11 UT on Sept 7 2026: the teal glyphs are Type III bursts firing with an M-class flare, next to the CME that erupted with it.
- Side view. A small edge-on inset showing how far above or below Earth's orbit each CME is aimed, which the flat map cannot show.

Side view. These Gannon CMEs straddle the ecliptic, which is why they hit.
- Top view. A small plan-view inset of the whole inner system. Most useful in 3D, and it stays zoomed out while you zoom the main map in.

Top view. The whole inner system at a glance, whatever the main map is doing.
- Sparkles. A drifting haze inside each CME. Decoration only, and it costs some speed on older machines.

Sparkles on. Eye candy, honestly labelled.
- Labels. All text on the map: planet names, CME speeds, distance rings and layer labels. Turn it off for a clean picture.
- Cinematic view. The default look: a textured Sun, a deep-space backdrop and glowing CME fronts. Switch it off for the original flat drawing. Same data and physics either way.

Cinematic (left) and classic (right). Same moment, same CMEs.
More detail
Fluid model: the solar wind as the moving fluid it is, painted across the whole plane under everything else: speed (blue slow, green and yellow faster, red 1,000+ km/s) or density (blue thin, red dense), switchable in the legend, on the same colour scales NOAA's Enlil pictures use. It's computed in your browser, in a couple of seconds, from two things this site already has: the wind we measured at L1 over the last solar rotation, mapped back to the Sun so the fast streams and slow sectors sit where they really were, and every CME on the map, launched as a dense fast pulse that plows through that wind. Fast wind catching slow wind piles up into the bright spiral bands; a CME's compressed sheath and the rarefied wake behind it form by themselves, because mass and momentum are conserved along each ray. The layer follows the playhead, Isolate and Solo like any other, so you can watch one CME's wake alone, or a whole week of eruptions interacting. Honesty notes, also in the legend: it's the ecliptic plane only, so a CME whose cone never crosses the plane is skipped and the legend says so; there is no magnetic field in it, so it says nothing about Bz or how strong a storm gets; densities are right in pattern, not calibrated in number; and the background wind is measured only where L1 had data, otherwise assumed. Tested the same way as the drag model, on 171 NASA-catalogued shocks from 2014 to 2024 where the CME's cone crosses Earth in the plane, its arrival-time error is 11.8 hours on average (median 9.2) against the drag model's 13.6 on the same events and NOAA Enlil's 11.2, using only wind measured before each CME launched; details in How accurate?. It is our own reduced-physics model after HUXt (University of Reading, MIT licence), not a NOAA product; NOAA's Enlil is a full 3D magnetohydrodynamic run on a supercomputer, and its images are the reference if you want the real thing.
More detail
Streams: high-speed streams, the fast solar wind (600 to 800 km/s) pouring out of coronal holes, drawn as the corotating spiral bands they really are, in teal, each labelled with its measured speed. They come from our own wind archive, not a model: a stream is drawn once the L1 wind has actually risen. And because coronal holes outlive a solar rotation, a dashed band marks where the same stream is predicted to return 27 days later. Live only: the layer clears during replays, where detection doesn't apply.
More detail
Parker spiral: the shape the Sun's magnetic field is dragged into by the solar wind: a garden-sprinkler spiral of faint field lines, wound tighter in slow wind and looser in fast. It's the reference grid for everything magnetic on this map: streams follow it, energetic particles ride it, and the Connectivity line is one strand of it.
More detail
Connectivity: shows the single Parker-spiral field line that links the Sun to Earth, marked at its solar footpoint (~W55 in typical wind). The footpoint is computed from the wind speed measured at the displayed moment, so it moves as you scrub or replay: slow wind winds the spiral tighter (footpoint near W65), a fast stream relaxes it (near W35). And when no measurement covers the displayed moment, the label says est. wind instead of pretending. A shaded green fan marks the well-connected zone, footpoint ±45°, because energetic particles travel along field lines: an eruption inside that fan reaches Earth in minutes, even when it's far off the Sun-Earth line. The second, brighter spiral that appears with a selection is the selected CME's own field line; it belongs to the selection, so it goes when you deselect (Esc, or ✕ on the card) and returns when you select again. Select a CME and its source is flagged connected (amber, particles stream to Earth) or not connected (dashed), with the angle from the footpoint. And when the eruption demonstrably dosed Earth, measurement outranks geometry: the line turns hot (☢ SEP observed) and shows how long the protons actually took to arrive. This is why a west-limb flare that misses Earth can still dose it (see the radiation card).
More detail
Spacecraft: marks where the observatories this site pulls data from actually sit. Three of them cluster at Earth: the L1 fleet (SOHO, ACE, SOLAR-1, IMAP, and DSCOVR, retired from real-time duty in 2026) sits ~1.5 million km sunward of Earth, only about 1% of the way to the Sun, so it hugs Earth until you zoom in. GOES and SDO orbit so close to Earth they're labelled on it rather than drawn separately. The interesting ones orbit the Sun independently: STEREO-A, well off the Sun-Earth line (a faint sightline shows its view toward the Sun), the second viewing angle that lets NASA triangulate each CME's true direction and width; Parker Solar Probe, diving through the corona itself; and Solar Orbiter, whose Venus-resonant orbit tilts higher with every flyby to climb toward the Sun's poles. Its label reads out its live latitude, the dimension the top-down map can't draw. Hover or tap any marker for a popup describing what that craft does. Solar Orbiter's popup also draws its orbit edge-on: the tilted ellipse against the flat ecliptic (Earth on the plane for contrast), its current height above or below it, and dashed guides at the planned 24° and 33° tilts of the coming Venus flybys. The layer is time-aware: a spacecraft only appears once the displayed time is past its launch, so a 2003 or 2012 replay shows only the craft flying then (no DSCOVR before 2015), and the 1859 Carrington replay shows an empty sky. Positions are approximate, plotted in the orbital plane.
More detail
Solar flares: draws each flare erupting at the Sun, timed to the playhead: a flare flashes when the timeline reaches its peak, then fades over about three days. It's drawn at the Sun's edge pointing in the flare's source direction, the same way the CME it launched points, and colored by GOES class (X red, M orange, C blue), with the strongest getting spikes. A dashed ring means that flare launched a CME; select that CME and its flare is highlighted. Play or scrub the timeline to watch flares fire in sequence.
More detail
Radio bursts: solar radio sweeps logged by NOAA's ground observatories, drawn at the Sun and timed to the playhead in teal so they read apart from flares: a Type II (a shock leaving the Sun) as a double arc, a Type III (electron beams) as spikes, others as a ringed dot. A Type II is the earliest sign a CME has launched: across 486 of them since 2015, 87% came with a catalogued CME, and the burst led NASA's first coronagraph sighting by a median 27 minutes (the analysis itself follows hours later). When we can pin the source to a flare, we draw a dashed provisional cone through the same physics as every other CME at the typical speed and width of Type II CMEs (670 km/s, 36°; half of them run 470 to 930 km/s). We tested deriving the speed from the burst's frequency drift on 393 events and it had no skill, so we don't. The cone is labelled "provisional" on the map, in the catalog and on its card, and dropped the moment DONKI posts the measured one. When the playhead crosses a burst, a faint ripple races to Earth's orbit. It starts where the burst came from and is brightest in that direction, because real bursts are beamed outward in a broad lobe, not sent equally everywhere. A located Type III also sends a dot out along the spiral magnetic field line from its source. That is the electron beam making the radio noise, at about a fifth of the speed of light, so it passes Earth's orbit about 45 minutes later. Radio takes eight minutes to get here, which no playback speed can show, so it's a cue, not a scale. On by default.
More detail
Side view: an edge-on inset (bottom-left of the map; the × in its corner turns the layer off) showing each cone's solar latitude: the dimension a top-down map can't show. A CME can look Earth-aimed in longitude yet fire well above or below the ecliptic. The side view is where you catch it. The Sun sits at the left, Earth at the right on the dashed ecliptic, and each CME is a slice tilted to its true latitude.
More detail
Top view: a plan-view inset of the whole inner system (the × in its corner turns the layer off). Most useful in the 3D view, where it keeps true heliocentric longitudes on screen while you fly the camera, but on the flat map it earns its place too: it stays zoomed out while you zoom in, so you never lose track of where Earth is.
More detail
Cinematic view: the app's default look: textured, slowly churning Sun, deep-space backdrop, glowing turbulent CME fronts, lit planets. Same physics, same data, same interactions. Only the rendering changes. Prefer the original flat rendering? Untick it; your choice is remembered, and ?viz=classic applies it via URL.
Time and playback
The tracker is a time machine: it can sit at the live moment, run forward to show forecasts, or rewind to replay what happened.
- The clock (top-left of the map) shows the displayed moment with a tag: LIVE (real time now), … AGO (the past), or +… FORECAST (the future).
- ▶ Play/pause animates time forward at the selected speed (also space).
- The two buttons either side of Play jump to the previous or next event (also [ and ]): a CME launch, an expected arrival at Earth, or an M or X flare. Landing on a launch or an arrival selects that CME, and a label says what you landed on.
- LIVE returns to live real time (also N or L).
- Speed buttons set how fast time runs:
1h/s,3h/s,6h/s,12h/s,1d/s(simulated hours-to-days per real second). - The timeline at the bottom is a scrubber spanning the loaded window. Triangles mark launches (coloured by speed class), dots mark predicted arrivals, and the orange line is the playhead; drag it to any moment.
- Isolate: the
Isolatebutton (right of the speed buttons) drops two handles on the timeline. Drag them to bracket a stretch of time, and the map replays only the CMEs that launched inside it. Handy for studying one burst of activity without the clutter of everything still in flight. Toggle it off to bring them all back. - Timeline span: the
30d / 60d / 90dselector at the right of the transport sets how far back the scrubber reaches. Selecting an older CME from the catalog auto-widens the span so the playhead always stays on the bar.

The timeline. Triangles are CME launches, coloured by speed class; dots at the top are arrivals, each joined to its launch. Right of NOW is the future: forecast arrivals.
The time charts
The small tab on the top edge of the timeline opens charts in a panel over the bottom of the map. Which charts it offers depends on the view: Time-distance and Storm anatomy in the Top-down, 3D and Magnetosphere views, and Satellite altitude in the Atmosphere view. Every chart follows the rest of the page: the same CMEs, filters, selection and playhead. Click a line to select that CME. Drag across a chart to move the playhead, exactly as you would on the timeline. Esc or × closes the panel.
Time-distance chart
Every CME is a line climbing from the Sun. Time runs across and distance runs up, so a steep line is a fast CME. Dots mark the forecast arrivals, each with a bar for its window, and green ticks mark shocks actually measured at L1.
More detail
Every CME is a line climbing from the Sun. Time runs across and distance from the Sun runs up, so the slope is the speed: a steep line is a fast CME. A fast line that bends over is the solar wind slowing it down, which is the drag model at work. The near-level lines are the planets at their true distance.
- Solid lines are CMEs aimed at Earth. Faint lines are aimed elsewhere. A faint line crossing Earth's line reached Earth's distance, not Earth.
- The dot on the planet's line is the forecast arrival, red for a direct hit and hollow amber for a glancing one. The bar around it is the honest window: the calibrated 80% window where the ensemble has run, and the model's typical error otherwise.
- Green ticks on Earth's line are shocks actually measured at L1. Where a green tick falls inside a bar, the forecast held.
- Lines that converge before Earth are CMEs catching each other up. The May 2024 replay shows several arriving as one pile-up.
10 daysframes the playhead.Whole timelineshows the full scrubber range.
Storm anatomy
One eruption from cause to effect on one time axis: the flare, the radiation, the CME in transit, the wind and Bz that reached L1, then Earth's response in Kp and Dst. Only the transit row is this site's model. Every other row is a measurement.
More detail
One eruption from cause to effect on a single time axis, top to bottom in the order things happen: the flare in X-rays, the radiation (protons), the CME in transit, the wind speed and Bz that reached L1, then Earth's response in Kp and Dst, and last the cosmic ray count, which dips when a CME's magnetic cloud shields Earth. With a CME selected the chart follows that eruption. With nothing selected it shows the last week. In a replay it spans the whole storm.
The transit row is this site's model. Every other row is a measurement. The shaded box is the forecast arrival window, drawn through the rows it predicts, so you can check it against the green lines where a shock was measured. Move the pointer over the chart to read every row at that moment.
Rows appear only when there is a record for the window. NOAA keeps seven days of X-ray and proton data, so older eruptions show flare peaks as stems and no proton curve. Replays carry flares, wind and Kp. The line under the chart names whatever is missing.
Satellite altitude
In the Atmosphere view the tab offers one chart: the measured altitude of the ISS, Tiangong and Hubble over 90 days, 1 year, or since 2015. Storms are shaded, and you can move the pointer over a line to read any day.
More detail
In the Atmosphere view the tab reads Altitude record since 2015 and opens one chart: the measured altitude of the ISS, Tiangong, Hubble and Swift, the same record as the drag card but larger and much longer: 90 days, 1 year, or Since 2015. Over the decade the ISS is a sawtooth of reboosts, Hubble is one long descent that steepens as the Sun climbs to maximum, Tiangong appears in 2021, and Swift falls from about 570 km to its reentry on November 4, 2026. Storms are shaded: every storm in the short view, only the strong ones (Kp 7 and up) in the long one. Move the pointer over a line to read that day. A drag moves the playhead within the last 90 days, which is as far back as the map goes. Changing to a view that does not offer the open chart closes it, and a ?chart= link opens a view that does.
Selected CME tab
Select a CME and the first tab on the right describes it, most important thing first.
- Verdict. Earth-directed, glancing, not Earth-directed, or not yet analyzed. Click Why for a drawing of the cone against the line from the Sun to Earth.
- Expected arrival. When the leading edge should reach Earth, with a countdown and the 80% window drawn as a bar. Where the ensemble has run, the time is its middle, the same one /today and the homepage give; the drag model's own time is noted under it, since that is when the map's front reaches Earth. A CME with no window yet shows the drag model's time, "give or take 12 h". While a CME is on its way, Email me when this one arrives signs you up for storm alerts, which cover every Earth-directed CME.
- Storm estimate and upper range. The likely geomagnetic level and a plausible upside. The gap is real: storm size depends on the direction of the CME's magnetic field, which is not known until it reaches L1.
- How the drag model's forecast has moved. Each shift of half an hour or more is kept, with the reason.
- How it turned out. Once the window has been closed for 60 hours, what actually arrived at L1 and how far it was from the forecast.
- The details. Launch speed and direction, the source region and flare, Solar Orbiter's viewing angle, transit time, the radio outlook, and NASA's note with a link to the DONKI record.
- Solo hides every other CME on the map. Copy link copies an address that opens the tracker on this CME.

An Earth-directed CME in October 2026: the verdict, the expected arrival and its 80% window, the storm estimate, how the forecast has moved, then the details.
Every line of the panel
- Verdict: Earth-directed, glancing, not Earth-directed, or not yet analyzed, with the angular offset from the Sun-Earth line. Click Why on that line for a drawing of the cone against the line from the Sun to Earth, with the angles that decide it: inside the cone is a direct hit, within 20° of its edge is a possible glancing blow, and anything further is a miss.
- Expected arrival: when the leading edge should reach Earth, with a countdown. Where the ensemble has run, the time is the middle of its 200 fluid-model runs, the number the site quotes everywhere; the single drag-model run's time is noted below when they differ by half an hour or more, because the map draws the drag model. Under it is the 80% window from the ensemble (see How accurate is it?), drawn as a bar from now to the end of the window with a tick at the most likely time. A CME with no ensemble yet shows "give or take 12 h", the model's typical miss. Once the window has passed, the block reads Forecast arrival and says how long ago. While a CME is still on its way, Email me when this one arrives signs you up for the storm alert list. That is the same list as everywhere else on the site, and it covers every Earth-directed CME, not only the one on screen.
- Storm estimate and upper range: the predicted geomagnetic response as two numbers, the most likely level and a plausible upside (e.g. Kp 4.4 and Kp 6.4). The gap between them is real: storm size mostly depends on the CME's magnetic-field orientation (Bz), which cannot be known until the cloud reaches the L1 monitor 30 to 60 minutes before Earth.
- How the drag model's forecast has moved: each time the drag model's arrival shifts by half an hour or more, the tracker keeps the old value and says why: NASA revised its measurement, or the solar wind changed. If NASA published a second measurement of the same CME, the panel also says where that one would put the arrival. Recording began on October 8, 2026, so older CMEs have no history.
- How it turned out: once the window has been closed for 60 hours, the panel says what the solar wind at L1 actually did: a front arrived inside the window, just outside it, or not at all, and how far it landed from the central forecast. This is the same scoring as the live scoreboard. When CMEs launched close together share one front, the panel says so.
- Speed & direction: the measured launch velocity and heading, and how the speed compares with the rest of the catalog.
- Source and source flare: the region it left from and the flare that launched it, if known.
- SolO view: the angle between the CME's axis and Solar Orbiter's vantage: near 90° means the spacecraft sees a clean side-on eruption, head-on means a hard-to-measure halo.
- Transit time and speed at Earth: how long the Sun-to-Earth journey takes, and how fast the front is still moving when it gets here.
- Radio: the HF outlook for that CME (degraded for 24 to 48 h after arrival, polar paths worst; possible auroral scatter on 6 m and 2 m), and any Type II or IV burst logged with its launch.
- NASA's note: the analyst's free-text comment from DONKI, followed by a link to the DONKI event itself.
- Solo hides every other CME on the map so you can watch this one alone, and it keeps working while pinned LIVE. Copy link copies a web address that opens the tracker on this CME (see Sharing and links).
The next four days
Under the selected CME, the same tab puts everything on its way onto one four-day strip: each Earth-directed CME's arrival window with a tick at the most likely time, any fast wind stream due back from a coronal hole, and NOAA's three-hourly Kp forecast with storm level marked. Click a CME's row to select it. Under the strip, Regions to watch lists the sunspot regions most likely to send the next CME toward Earth, with NOAA's flare odds for the day.
How the region odds are worked out
Each region's number is its chance of producing an Earth-directed CME in the next three days, computed here from where it sits, its magnetic class and area, and what it did in the last week. Only regions within 45° of the Sun-Earth line, or about to rotate into it, can score much. Two honesty rules are built in. The odds are capped at 30%, because above that the model over-forecast in testing. And about half of all Earth-directed CMEs come from no numbered region (filament eruptions, stealth CMEs, sources NASA never pinned down), so an empty list never means nothing is coming. The model is a nine-term logistic fit on 14,175 region-days from 2014 to 2024 (scripts in the repo): a region inside the zone that fired in the last week reaches an Earth-directed CME within three days about 19% of the time, a beta-gamma-delta region that fired about 38%, against 3% for a region outside the zone. Cross-validated discrimination is solid (AUC 0.73 to 0.76) and the reliability curve is straight below 30%. Because NASA attributes far more CMEs to regions now than in 2016, the model's baseline is re-solved monthly on the trailing year of outcomes rather than trusted from the original fit.
Inbound CME warning
When a CME is genuinely en route and worth watching, a pulsing banner appears over the map. It fires when a storm of G2 or stronger is plausible: the upside of the calibrated estimate range, not the midpoint, because that's the watch-worthy number. If several qualify, it shows the strongest, then soonest.
The banner carries a watch-style severity chip ("≤G2 MODERATE", the plausible ceiling), the impact time, a countdown, and the estimated Kp range (e.g. "Kp 4.8 to 6.8"); its color and pulse rate scale with that ceiling. Dismiss it with × and it stays hidden until a different CME becomes the top threat.
Earth tab
What the Sun is doing to Earth: the globe, the magnetosphere, Kp, satellite drag and radiation.
The globe
A globe rendered from the Sun's point of view, so you see which face a storm would strike on arrival. It updates to whatever moment the clock shows. In cinematic view it renders real NASA Blue Marble imagery (re-projected to the sub-solar hemisphere) with aurora curtain rays; classic view keeps the original drawn globe.

The Earth globe: the sunlit face at the displayed moment, with NOAA's aurora oval.
More detail
- Day/night terminator & sub-solar point: the lit hemisphere and the spot where the Sun is directly overhead. This is how the tool knows which side of Earth is sun-facing when a CME arrives.
- The magnetosphere: the blue field bubble. It compresses when solar-wind dynamic pressure rises, and energy couples in when the field turns southward (negative Bz).
- Aurora ovals: drawn around both poles. At the live moment they are NOAA's OVATION nowcast. At any other moment they are sized by the Kp index, and high Kp pushes them toward the equator.
The small tag by the title states what you're seeing: live now, at map time, ⚡ CME impact, ✓ Earth clear (a near-miss), replay, or at CME arrival. The caption beneath spells out the sub-solar point, the aurora Kp, and the wind/Bz coupling.
Who drives the globe: selecting a CME with an Earth arrival pins the globe to that arrival moment while playback is paused (the at CME arrival preview: which face gets hit). The moment you move the timeline yourself, by drag or arrow keys, the globe follows your playhead instead, and stays with you until you select a CME again, which re-arms the preview.
Magnetosphere
The card draws Earth's magnetic shield as the solar wind is shaping it at the displayed moment. The headline number is the distance from Earth to the front of the shield in Earth radii: about 10 in quiet wind, less when the wind is dense, fast or southward. The card turns red when the shield is pushed inside the orbit of geostationary satellites. Click the card, or pick Magnetosphere in the view buttons, for the full view. There the shield is a glowing surface, the solar wind streams in from the Sun and bends around it, and a dashed outline shows where the shield sits on a quiet day, so a storm reads as the gap between the two.
More detail
The Magnetosphere card on the Earth tab draws Earth's magnetic shield as the solar wind is shaping it at the displayed moment: the magnetopause in amber, where the wind's pressure balances Earth's field, and the bow shock in cyan standing off ahead of it. The Sun is to the left and north is up. The headline number is the distance from Earth to the nose of the magnetopause in Earth radii; about 10 in quiet wind, and pushed inward by dense, fast wind and by southward Bz. The dashed line through Earth is geosynchronous orbit seen edge-on, at 6.6 Earth radii: when the nose comes inside it the card turns red, because the satellites on the day side are then sitting out in the solar wind. That happened during the Gannon storm; scrub the replay to May 10 2024 and watch it. Click the card, or pick Magnetosphere in the view control, for the full rotatable view on the map stage, driven by the timeline like everything else. The shapes are the Shue et al. (1998) magnetopause and the Farris & Russell (1994) bow-shock stand-off, empirical average surfaces fit to decades of spacecraft crossings. They are axisymmetric, have no cusps or dawn-dusk asymmetry, and the real boundary wobbles around them by an Earth radius or so; treat the number as a good estimate, not a measurement. The L1 confirmation email quotes the same figure. The view opens from the side, with the Sun to the left and marks every 5 Earth radii along the Sun line. Drag to rotate and scroll or pinch to zoom, or pick From above, Free orbit or Close to Earth from the camera select beside the view control. Two lines at the bottom left give the reading in words: how far out the nose of the shield is, and whether geostationary satellites are inside it. The streaks are the solar wind: they move at a pace set by the measured speed, there are more of them when the wind is dense, and they slow and turn where they meet the bow shock. They illustrate the flow and are not a simulation of it. The dashed grey outline is the same shield in ordinary wind, about 11 Earth radii out, drawn whenever the present shield differs from it. Close to Earth, the planet is shown with its real day and night sides and the aurora ovals ring the magnetic poles over real geography, wider and brighter as the storm builds. Earth-directed CMEs appear on the stage as a bright front in their speed-class colour that closes in from the Sun side over the 12 hours before their forecast arrival, then sweeps through and fades. At this scale a CME front really is a flat wall, but its approach is drawn as a countdown rather than to scale: at 600 km/s it would cross the whole view in seven minutes. With the Spacecraft layer on, GOES-East, GOES-West and SDO sit on the GEO ring at their real local hour angle and turn red when the boundary has left them out in the solar wind, and a marker shows where the L1 fleet is, 235 Earth radii up the Sun line. The Side view and Top view insets draw here as well.
Geomagnetic activity (Kp)
The planetary K-index in 3-hour steps. Kp ≥ 5 is a geomagnetic storm; the bars are colored by severity.
Satellite drag and the Atmosphere view
Storms heat the upper atmosphere. It swells, and everything in low orbit meets denser air and sinks faster. The card's headline is air density at 400 km compared with a quiet day: about 1 when calm, 2 to 3 in a severe storm. The Atmosphere view draws the atmosphere in shells with the ISS, Tiangong, Hubble and Starlink at their real altitudes. A dashed ring marks where the air of 400 km sits on a quiet day and a bright ring marks where that same air is now, so the swelling shows as the gap between them. The satellites circle their rings, each trailing the path it has sunk along: the faster it is sinking, the steeper the trail. Rings, orbits and trails are not to scale. A satellite turns red when it is sinking at least 2.5 times its quiet rate. The green lines on the globe are the aurora zones, where a storm's energy enters. They brighten and move toward the equator as Kp rises, and an orange glow spreads from them while the storm is heating the air. The glow is the model's picture of where the heat goes, not a measured map. The list at the top left gives each satellite's sinking rate from the model, with the measured rate from tracked orbits beside it. The measured rate jumps around from day to day, so the two rarely match exactly. A link under the list opens the measured altitude record back to 2015. Under the model, the card shows what the three stations actually did, day by day.
More detail
The Satellite drag card and the Atmosphere view take the storm one step further down: into the upper atmosphere. Geomagnetic storms heat the thermosphere, it swells outward, and everything in low orbit meets denser air and sinks faster. The view draws the atmosphere as altitude shells coloured by air density, brighter and warmer as it inflates, with the ISS, Tiangong, Hubble, NASA's Swift observatory and the Starlink shell at their real altitudes from the day's orbital elements and the decay each one implies. Swift had no engine and reentered on November 4, 2026. It is gone from the scene after that date and comes back when you scrub to an earlier day. The Swift page tells the whole fall. The headline is air density at 400 km as a multiple of a quiet day: about 1 in calm conditions, 2 to 3 in a severe storm, and it stays high for a day or two after the storm ends because the atmosphere cools slowly. A second line says how long a batch launched to 210 km would last, which is the February 2022 Starlink loss in one number. The model is a simple scale-height atmosphere driven by F10.7 and the previous three days of geomagnetic activity, not NRLMSIS. Its constants were fitted against eleven years of measured decay for the ISS, Tiangong and Hubble from their published orbital elements, so its storm response matches the record, and where we have a day of measured decay for an object it sits next to the model's figure as the ongoing check. The live scoreboard keeps the running tally: over the last 90 days, how far the model's decay sits from the measured decay for the ISS, Tiangong and Hubble, and what the ISS actually did on storm days against what the model said.
Measured altitude. Under the model, the card draws what the four objects actually did: the daily altitude of the ISS, Tiangong, Hubble and Swift from their orbital elements, over the last 90 days, or around the storm in a replay. Each has its own panel and its own scale, because the point is the shape of each line. A steady slide is ordinary drag. A dashed riser with a ▲ is a reboost. A step down through a shaded band is a storm heating the upper air. The strip under each line lights on days the object sank at least twice its usual rate. Hubble has no engine, so its line only ever goes down.
One-day differences of orbital elements are noisy: a quiet day for Hubble reads anywhere from 0 to 60 m. So the sentence under the chart credits a storm only when a drop is large, falls next to a storm, and a second object sank fast on the same day. Small storms do not show, and the sentence says so. The May 2024 replay is the clear case: all three drop on May 11. The orbit archive starts in 2015, so older replays show the model alone.
Radiation belts
The Radiation belts layer draws the Van Allen belts in the Magnetosphere view. The outer belt's brightness follows the electron flux GOES measures at geostationary orbit, and the card says when that flux is above NOAA's alert level.
More detail
The Radiation belts layer (on by default) fills in the Van Allen belts as the space between the field lines: the inner proton belt, steady, and the outer electron belt, whose brightness follows the GOES ≥2 MeV electron flux at geosynchronous orbit that NOAA publishes every five minutes. When a storm pushes the magnetopause in, the outer shells that reach the boundary drain, and in the days after, the belt refills hotter with its peak moved inward; the card says when the flux at GEO is above NOAA's 1,000 alert level, the regime where satellites charge internally. The belt shape is an illustrative profile pinned to the GOES value, not a radiation model, and our electron archive starts in September 2026, so earlier replays draw no belts.
Radiation storm (S-scale)
Solar proton flux on NOAA's S1 to S5 scale (set by the ≥10 MeV flux crossing 10 pfu). It's the one space-weather hazard with a direct radiation dose to people (astronauts, polar-route aircrew), so the card names who's affected at each level. The top shows current conditions live, or that storm's peak in a replay. Select a CME and a line appears for the radiation storm that eruption produced: protons arrive within minutes-hours of launch (not days later with the CME), so a flare can dose Earth long before its CME arrives. And even a CME that misses Earth can dose it via magnetic connectivity.
Sun tab
Imagery from NASA's Solar Dynamics Observatory: live when the playhead is at "now" (refreshed every few minutes), and time-synced to the playhead otherwise: scrub or play into the past and the panel shows the closest archived full-disk frame (hourly, via Helioviewer), with a timestamp chip marking the moment. Frames are cached as you go, so replaying a storm gets smoother the more you watch. Before SDO's first light (May 2010) a placeholder appears instead. Three views:
- Corona 193Å: the million-degree corona; coronal holes and active regions.
- Chromo 304Å: the cooler chromosphere; prominences and filaments.
- Sunspots: the visible-light photosphere with sunspot groups.
Recent flares pulse at their location on the disk, timed to the playhead so they match the frame you're looking at, and numbered active regions (sunspot groups) are marked when the image is current. The caption summarizes the current regions and flare tally.

The Sun panel. SDO imagery at the displayed time, with today's numbered regions ringed.
X-ray flux
The card under the Sun panel draws the last three days of GOES X-ray flux, the measurement that flares are classed by. The scale is logarithmic: each letter (A, B, C, M, X) is ten times the one below it, so an X1 flare is a hundred times a C1. The shaded band at the top is M and X, where a flare is strong enough to black out HF radio on the sunlit side of Earth. The largest flares are named at their peaks, and the orange line is the playhead.
More detail
In a replay the card draws each catalogued flare as a stem at its peak class instead, because the continuous record is not loaded for replays. The card fetches nothing until you scroll to it.
Coronagraph
The card under the Sun panel shows the outer corona from NOAA's two coronagraphs: CCOR-2 on SOLAR-1 at L1 (the default) and CCOR-1 on GOES-19 in Earth orbit. This is the view every launch speed in the catalog comes from: a CME leaving the Sun appears as a bright loop pushing outward from behind the occulting disk. Like the Sun panel it is time-synced to the playhead, served from this site's own archive of frames (every frame since the instruments came online, plus older CCOR-1 frames back to 2025), with a timestamp chip and a note when the nearest frame is more than an hour away.

The coronagraph card. NOAA's CCOR imagery, scrubbable in time like the rest of the page.
Running difference
The Δ Diff button on the coronagraph card shows each frame minus the one about 30 minutes before it. The steady corona is the same in both frames, so it cancels and only what changed is left. Grey means nothing changed. Bright means the corona got denser there. Dark means it thinned. A CME shows as a bright arc moving outward with a dark wake behind it, and a faint one that is hard to find in the plain frame is usually obvious here. This is the same technique forecasters use to measure a CME's speed.
More detail
The subtraction runs in your browser on the frames the card already holds. The frame label shows the gap that was used, for example Δ 30 min. If there is no earlier frame to subtract, the card shows the plain frame and says so. Small speckle is image noise, not the Sun.
Solar wind tab
What is being measured right now at L1, the point about 1.5 million km toward the Sun where spacecraft sample the wind 30 to 60 minutes before it reaches Earth.
- L1 nowcast: the highest-confidence short-range read. Because L1 sits ~1.5 million km sunward, whatever it measures reaches Earth ~30 to 60 min later (sooner when the wind is fast). It's a measurement with a travel time, not a model. When quiet it shows the current wind, Bz, and that travel time; when a shock crosses L1 it flags an incoming sudden commencement with a live countdown to Earth; sustained strongly southward Bz warns that energy is about to couple into the magnetosphere. A nowcast, not a forecast: it reports only what's been measured.
- Solar wind · 7 days: wind speed and the interplanetary magnetic field Bz. Bz turning southward (negative) is the key storm trigger. It lets solar-wind energy pour into the magnetosphere.
The 3D view
The view control beside the Layers menu has four buttons: Top-down, 3D, Magnetosphere and Atmosphere. 3D swaps the flat map for the same solar system rendered as an actual three-dimensional scene. Everything else is unchanged: same physics, same catalog, same timeline, same clicking and hovering, and both the cinematic and classic looks apply. Only the projection differs.
Why it exists: latitude. The flat map draws every CME from its measured longitude and half-angle, and has no way to show the third number NASA also measures: the cone's latitude. So a CME erupting 55° above the Sun's equator gets painted sweeping straight across Earth, and then the card calls it a miss. In 3D the cone is simply where it really is: sailing over Earth's head. That contradiction is the best single reason to switch views.
Moving the camera
- Drag to orbit the Sun; scroll or pinch to move closer or further out. The reset button (the circular arrow) puts the camera back.
- The camera never rolls, and it stops just short of the poles, so the plane of the solar system always reads level however far you swing around. Look straight down and you are back to very nearly the flat map, which is what makes moving between the two views readable rather than disorienting.
- The camera travels. Switching between Top-down and 3D is one continuous move: the camera rises to look straight down, the flat map takes over, and the other way round. Changing vantage point flies you there. If your device asks for reduced motion, it cuts instead.
Vantage points
The dropdown beside the view buttons decides where the camera sits.
- Free orbit, the default: you float outside the system looking in.
- From above: the flat map's own angle.
- Edge-on: level with the planets, which shows how far north or south each CME is aimed.
- Follow CME: the camera rides with the selected CME.
- From Earth: looking back at the Sun, the way a coronagraph sees it. A CME aimed at you looks like a ring opening around the Sun, which is called a halo.
More on each vantage point
The dropdown to the right of the view switch decides where the camera sits. (It is hidden in top-down mode, which has no camera to place.)
- Free orbit, the default: you float outside the system looking in.
- From above: the flat map's own angle, as a starting point. The first drag turns it back into free orbit.
- Edge-on: level with the plane of the planets and side-on to the Sun-Earth line, with Earth to the right. This is the angle that shows how far north or south of us each CME is aimed.
- Follow CME: the camera rides with the selected CME and keeps it centred while it travels. With nothing selected it follows the next front due at Earth. Drag to swing around it.
- From Earth: the camera stands on Earth and looks back at the Sun. This is the geometry a coronagraph sees, and it is worth understanding: a CME aimed at you does not look like a cone off to one side, it looks like a ring opening around the Sun. That is a halo CME, which is why halos are at once the dangerous ones (they are coming at us) and the hardest to measure, since a ring hides how fast it is really travelling and looks much the same whether it is heading toward you or directly away. Here you can watch a wedge become a halo just by changing seats. In this vantage, dragging looks around and scrolling changes the focal length rather than the distance.
- When a front finally sweeps over your vantage point, the frame washes in that CME's colour: the storm arriving, seen from inside it, at the transit time the model predicts.

The 3D view with one CME soloed, mid-flight. The chip at the nose reads distance, latitude and speed; the dashed tether drops to the orbital plane so the latitude is visible; the compass shows the camera tilt.
What else is drawn differently in 3D
- Earth pulses on impact on exactly the same frame as it does on the flat map. It is the same arrival test, not a second animation that might disagree with it.
- A CME is drawn as a whole cone. The bright cap is the front. Two faint sides run back to the Sun so you can see where it came from, and nearer fronts are drawn heavier than far ones.
- Every cloud has a footprint. Its outline is dropped straight down onto the plane of the planets, with a dotted line from its nose. A cloud whose nose stands well clear of its footprint is passing above or below the planets.
- Old fronts retire. On a busy day forty-odd cones are in flight at once, and the oldest are also the largest, so left alone they bury the Sun and planets entirely. Once a front has passed Earth it thins to a bare ring and then leaves the picture, because from that moment it is history. The selected CME is never retired.
- Speeds are labelled only where they matter: on fronts aimed at Earth, and on the selected CME.
- The Sun is the real Sun. Once it is big enough on screen to show detail, its surface is the SDO image for that hour, in the channel the Sun panel is set to, wrapped onto the ball. From Earth it is exactly that image. The side SDO cannot see is left plain and darker. With a CME selected, a ring marks where on the Sun it came from, carried along as the Sun turns.
- The forecast's error bar is drawn in space. For the selected CME, two dashed rims marked early and late show where the front would be right now if it arrives at the early or the late edge of its arrival window. The gap between them is the uncertainty.
- The solar wind lies on the plane. Turn on the Fluid model and its speed or density map becomes the ground the planets sit on. Streams and the Parker spiral are drawn on the plane as well.
- Flares flash where they happened. A solar flare is drawn on the Sun's surface at its own latitude and longitude and turns with the Sun. One on the side facing away from the camera shows only the ring around the Sun at the moment it fires.
- Connectivity lies on the plane. The field line linking the Sun to Earth, its footpoint, the well-connected zone and the selected CME's own field line are drawn on the plane of the planets, with the same colours and labels as the flat map.
- Every map layer works here. Nothing in the Layers menu is greyed out in 3D.
- Spacecraft is worth turning on here. Parker Solar Probe and Solar Orbiter both fly tilted orbits, and this is the only view that can draw them where they actually are instead of reporting their height as a number in a label. Solar Orbiter swings up to roughly 0.18 AU clear of the plane, which is the entire point of the mission.
- Side view and Top view both work here, as picture-in-picture panels stacked at the bottom-left. The top view is the valuable one in 3D: perspective is what makes a three-dimensional scene worth having, and also what makes it worse at the question the flat map answers instantly: how far off the Sun-Earth line is this thing? Keeping the plan view in the corner gives you both answers at once.
Which view you get
The tracker opens top-down. Once you pick a view yourself, your browser remembers it. ?view=3d and ?view=2d set it from a link, and &eye=earth picks the vantage. Replays are a good place to try 3D: they carry about twenty CMEs, not several hundred, and they are the storms where a cone's true latitude matters.
Replaying a past storm
Opening any storm from Famous CMEs loads it and turns the whole app into a scrubbable replay of it. Great for seeing extremes and sanity-checking the model against known events. A banner shows you're in replay; ↩ Live returns to real time. Replays are also the best place to try the 3D view, where a storm's true latitude is visible. Each one's full story is on Famous CMEs: the history, the impacts, and the science.
| Event | Class | Why it matters |
|---|---|---|
| Gannon · May 2024 | G5 | Biggest storm in 20 years; aurora to the tropics. (DONKI-measured CMEs.) |
| September 2017 | G4 | X9.3, the largest flare of the solar cycle. (DONKI-measured CMEs.) |
| St. Patrick's · 2015 | G4 | A modest CME that punched far above its weight. (DONKI-measured CMEs.) |
| 2012 Near-Miss | none | A Carrington-class CME that crossed Earth's orbit and missed. (Measured at STEREO-A.) |
| Starlink Storm · Feb 2022 | G1 | A minor storm whose atmospheric drag destroyed 38 satellites. (DONKI-measured CMEs.) |
| Halloween · 2003 | G5 | The X17/X28 superstorms; grids and satellites hit. (CME reconstructed, measured wind + Kp.) |
| Quebec Blackout · 1989 | G5 | Grid collapse in 92 seconds; 6M people dark. (Measured Kp; wind & CME reconstructed since the storm fell in a data gap.) |
| Bastille Day · 2000 | G5 | Extreme Bz (−60 nT); aurora to Texas. (CME reconstructed, measured wind + Kp.) |
| Carrington · 1859 | G5 | The most intense storm on record. A reconstruction (no instruments existed). |
| Nine-Night Storm · 1770 | Carrington-class | Five CMEs in a week from a sunspot group twice Carrington's size; low-latitude aurora nine nights out of ten. (Reconstruction.) |
| Railroad Storm · 1921 | G5-class | The 20th century's deepest storm; telegraph fires on two continents. (Reconstruction.) |
| Almost-War · May 1967 | Kp 9 | The flare that jammed missile-warning radars and nearly started a war. (Kp measured.) |
| Sea-Mine · Aug 1972 | Kp 9 | Fastest transit on record (~14.6 h); detonated sea mines off Vietnam. (Kp measured.) |
Provenance is labeled honestly: measured archive data vs. reconstruction (estimated). For the 2012 near-miss, a special card shows what STEREO-A measured, the storm Earth was spared, while the globe correctly stays calm.
The replay banner. Everything on the page below it is the archived data for that storm; Live returns to now.
Each replay has its own web address. See Sharing and links.
On a phone, and as an app
The tracker restacks on a phone: header readouts (wind, Kp and Dst), the map with Layers and the view switch, the transport and timeline, then the catalog and the dashboard cards in a single column.
- Scroll down to the cards and a slim time bar pins itself to the bottom of the screen: play, the clock, a scrubber and LIVE. It lets you move through time while looking at the Earth globe, the Sun image or the coronagraph, which all follow the clock. It goes away when the timeline is back on screen.
- The catalog's filters fold behind a ⚙ Filters button, and the list is capped behind Show all CMEs.
- Tap a wedge or a row and the Selected CME card opens as a sheet over the map. Drag the handle at its top up or down to make it taller or shorter, swipe it down to close it, or tap the ✕.
- The view switch and the camera select merge into one selector: Top-down, 3D (free orbit, from above, edge-on, follow a CME or from Earth), Magnetosphere (from the side, from above, close to Earth or free orbit) and Atmosphere.
- In the Atmosphere view, tap a satellite's dot for its numbers and tap elsewhere to clear them. The density bars sit under the Earth.
- Some desktop tools are left out for space: Isolate, the 30/60/90-day span, the UT toggle, PNG export, clip recording, the 3 h and 12 h speeds and the Recent flares list. Rotate to landscape or use a tablet for the full set.
- Hover tooltips don't exist on touch; the help you'd get from them is on this page.

The phone layout: the map first, the timeline under it, and the panels stacked below.
Install it as an app
CME Tracker can be installed like an app. It gets its own icon and its own window with no browser bars, and it opens straight to the live tracker. It is the same site, so there is nothing to update and nothing comes from an app store.
- Chrome or Edge on a computer: use the button above when it shows, or the install icon at the right end of the address bar.
- Chrome on Android: use the button above, or the browser menu, then Add to Home screen.
- iPhone or iPad: open the site in Safari, tap Share, then Add to Home Screen.
- Safari on a Mac: File, then Add to Dock.
The app needs a connection, because everything it shows is live. To remove it, uninstall it the way you would any app.
How accurate is it?
The arrival forecast is back-tested. Every arrival the model would predict is compared against NASA's catalog of observed shock arrivals (hundreds of real CME→shock pairs); the full scoreboard and error histogram live on the science page.
- Typical (median) error ≈ 12 h; mean ≈ 15 h.
- Essentially unbiased: it doesn't systematically run early or late.
- That test is generous: it uses hindsight and NASA's final measurements. Scored the way a real forecast would be, the mean miss is about 17 h and the median about 13 h, and half of arrivals fall outside the "±12 h" shown on the card.
- NASA's full MHD model (WSA-Enlil) is better. On the storms both forecast, its first real-time run missed by about 12 h on average against our 16 h. Expected, since Enlil resolves the structured solar wind that a two-parameter drag model only approximates.
- Whether a CME arrives at all is a separate question. Over 2014 to 2024 a DIRECT call was right about six times in ten, and it caught 58% of the CMEs that arrived. The hit and miss table has the full record.
- The big misses are usually slow or poorly observed CMEs. The histogram on the science page shows the spread.
Live scoreboard. Every arrival window the site issues is scored against our own L1 archive once it has closed: did a front arrive inside it, just outside it, or not at all, and how far from the forecast time. Two kinds of front count: shocks (a 30 km/s jump with density and field both rising; that catches about 57% of the fronts NASA lists, the clear ones, with 81% precision) and gradual arrivals (a climb of 80 km/s or more inside six hours that falls back within a day and a half, the signature of a CME body reaching L1 without a shock, which is how the Sept 6 2026 CME arrived; a rise that stays up for days is a coronal-hole stream and is not counted). Loading the record… The chart of every forecast against its arrival, and the scoring rules, are on the science page.
How the window, the storm estimate and the model work
The arrival window on the card comes from an ensemble. For every Earth-ward CME, a cron runs the fluid model 200 times with the cone's speed, direction, width and launch time and the background wind each perturbed by their known errors, then dresses each run with the model's own error, calibrated so the 10th-to-90th-percentile window really contains 80% of the catalogued arrivals (it does: 79% on the 214 rows where the cone crosses Earth, at a typical width of about 38 hours, or ±19). That's wider than the "±12 h" typical error above because it is an honest 80% window, not a median. Only wind measured before each CME launched goes in, so nothing about the storm leaks into its own forecast. The share of runs that cross Earth in the ecliptic is shown too, but it can't be calibrated on a catalogue that only contains hits, so read it as "how close to the cone's edge Earth sits", not as a probability of impact.
Storm magnitude has its own back-test. The Kp estimate is calibrated against 221 observed CME→storm events (2014 to 2024; the subset of the 282 timed arrivals that came with a clean measured storm response): the most-likely level is right or within one G-level about 84% of the time. That number flatters the estimate. Fitted on 2014 to 2020 and tested on 2021 to 2024, it scored the same as saying "Kp 4.3" every time (the test). At launch it says what CME storms are usually like and almost nothing about this one. At a fixed prediction, real outcomes still span roughly ±2 Kp, because the deciding variable (the CME's field orientation) is unknowable until L1. That's why everything here shows a range rather than a single confident number; a single-level call only becomes honest ~30 to 60 minutes out, when the field is actually measured.
The model is the analytic Drag-Based Model (Vršnak et al. 2013): each CME starts at 21.5 solar radii with its measured speed and coasts toward the ambient solar-wind speed under aerodynamic drag. Most of the remaining error comes not from the propagation but from the input: the coronagraph-derived speed and direction, especially for halo CMEs.
Want the full derivation? The science & math behind CME Tracker walks through every equation: the drag model, the cone geometry, the ephemerides, the storm heuristic, and the validation methodology, with the limitations stated plainly.
Controls
Keyboard
| space | Play / pause time |
| N / L | Return to live / exit replay |
| ← / → | Step time back / forward (hold shift for a full day) |
| [ / ] | Jump to the previous / next event (launch, arrival or M/X flare) |
| ↑ / ↓ | Select previous / next CME |
| 1 / 2 / 3 | Layer presets: Monitoring / Clean / Sensory overload |
| Esc | Close a popup, the Layers menu or a dialog; with nothing open, deselect the CME |
| Enter | Open the popup of a focused header readout |
Mouse
| Scroll | Zoom the map · in 3D, move the camera nearer or further |
| Drag | Pan the map · in 3D, orbit the camera (or look around, from a body vantage) |
| Click cone / row | Select a CME |
| Hover | Quick tooltip |
Other tools
UTtoggles between UTC and your local time zone everywhere (display only, no change to the physics).- Collapsible panels: the chevron tabs (
‹/›) at the left and right edges of the map hide the catalog or the dashboards, expanding the map to fill the space. Your choice is remembered between visits. - The
Layersmenu groups the map overlays: Fluid model, Magnetic wake, Streams, Parker spiral, Connectivity, Spacecraft, Radiation belts, Solar flares, Radio bursts, Side view, Top view, Sparkles, Labels and Cinematic view. Each is described in Layers. - The view control (
Top-down,3D,Magnetosphere,Atmosphere) changes what the stage shows, and the camera select beside it places the camera: Free orbit, From above, Edge-on, Follow CME or From Earth in 3D, From the side, From above, Close to Earth or Free orbit on the magnetosphere stage (see The 3D view and Magnetosphere). - Collapsible cards: click any card title in the right panel to fold it to one line. A folded card shows a live summary in its header. A card whose state turns urgent (a shock or strongly southward field at L1, a radiation storm, the magnetopause inside GEO, storm-level drag, Kp 5 or more) opens itself once; close it and it stays closed until that episode ends. Your folded set is remembered.
Isolatereplays just a chosen stretch of the timeline (see Time and playback).- Live feeds refresh automatically every few minutes; the status line at the right of the header says when the data was last updated.
Glossary
- CME
- Coronal mass ejection: a large eruption of magnetized plasma from the Sun.
- Solar flare
- A sudden burst of radiation from the Sun, classed A/B/C/M/X by X-ray brightness. Often (not always) accompanies a CME.
- Solar wind
- The continuous stream of charged particles flowing out from the Sun (~300 to 800 km/s).
- IMF / Bz
- The interplanetary magnetic field carried by the solar wind. Its north-south component, Bz, is decisive: strong southward (negative) Bz drives geomagnetic storms.
- Kp index
- A 0 to 9 scale of global geomagnetic disturbance, in 3-hour steps. Kp ≥ 5 = storm.
- NOAA G-scale
- Storm severity from the peak Kp:
| Scale | Kp | Effect |
|---|---|---|
| G1 Minor | 5 | Minor grid swings; aurora at high latitudes. |
| G2 Moderate | 6 | Aurora to mid-latitudes; HF radio fades. |
| G3 Strong | 7 | Surface charging; satellite drag; wider aurora. |
| G4 Severe | 8 | Grid voltage problems; aurora to low latitudes. |
| G5 Extreme | 9 | Grids at risk; aurora near the tropics. |
- Half-angle
- Half the angular width of the CME cone, as fitted from coronagraph imagery. The map draws each cone spanning ± the half-angle around its axis.
- Halo CME
- A CME aimed nearly along the Sun-Earth line, so it appears as a halo around the Sun in coronagraphs. Usually Earth-directed.
- R☉ / AU
- Solar radius (~696,000 km) and astronomical unit (~150 million km, the Sun-Earth distance). CMEs are tracked from 21.5 R☉ outward.
- DBM
- Drag-Based Model: the analytic propagation model used here (see How accurate is it?).
- L1
- The Sun-Earth Lagrange point ~1.5M km sunward, where SOLAR-1 (with ACE and IMAP as backups; DSCOVR until spring 2026) measures the wind ~30 to 60 min before it reaches Earth.
- Magnetosphere
- The protective magnetic bubble around Earth; the magnetopause is its sunward boundary, which compresses under a storm.
- Sub-solar point
- The point on Earth where the Sun is directly overhead.
- Dst
- Disturbance storm-time index, in nanotesla: how much the ring current circling Earth has weakened the field at the equator. More negative is a deeper storm; −100 nT is intense, the Gannon storm reached −412, Carrington is estimated near −900. Hourly, from Kyoto.
- Coronagraph
- A telescope that blocks the Sun's disk with an occulting disk so the faint outer corona, and a CME leaving it, becomes visible. Every launch speed and direction in the catalog starts as a coronagraph measurement.
- Coronal hole / high-speed stream
- An open-field region of the corona that lets fast wind (600 to 800 km/s) escape. The stream sweeps past Earth as a corotating spiral and returns about 27 days later; see the Streams layer.
- Type II radio burst
- A slow-drifting radio sweep from a shock leaving the Sun, usually the first sign a CME has launched, typically half an hour before a coronagraph sees it.
- Sheath / flux rope
- The two parts of a CME at Earth: the compressed, turbulent wind piled up behind the shock (the sheath), then the smooth twisted magnetic cloud itself (the flux rope), whose Bz decides the storm.
- Aurora oval
- The ring around each magnetic pole where aurora occurs. Kp sets how far toward the equator it reaches; the Earth globe draws it at the displayed Kp.
Frequently asked questions
What is a coronal mass ejection (CME)?
A coronal mass ejection is a large eruption of magnetized plasma, often a billion tons of it, blasted off the Sun's corona. When an Earth-directed CME arrives, typically 1 to 3 days after it launches, it can compress Earth's magnetic field and trigger a geomagnetic storm, auroras, and disruptions to satellites, radio, and power grids.
How accurate are CME arrival-time predictions?
CME Tracker's arrival estimates have a median error of 12.3 hours and a mean of 14.8 hours, validated by back-testing the model against 282 real shock arrivals in NASA's catalog (2014 to 2024). NASA's full magnetohydrodynamic model, WSA-Enlil, scores a mean of 11.2 hours on the same events, so it is better; both are hours, not minutes. Most of the remaining error comes from the coronagraph-measured launch speed and direction, not the propagation math. Treat arrivals as indicative, not operational forecasts.
How does the model predict when a CME will hit Earth?
It uses the analytic Drag-Based Model (Vršnak et al. 2013). Each CME starts at 21.5 solar radii with the speed NASA's coronagraphs measured, then decelerates (or accelerates) toward the ambient solar-wind speed under aerodynamic drag, so fast CMEs lose much of their speed before arrival. A 3,000 km/s CME can arrive at under half that speed.
How much warning is there before a solar storm hits Earth?
It depends on the source. A CME's arrival is forecast 1 to 3 days ahead (±~12 h). The highest-confidence warning, though, comes ~30 to 60 minutes out: spacecraft at the L1 point (~1.5 million km sunward) measure the solar wind and magnetic field before it reaches Earth, so when a shock crosses L1 the tracker's L1 nowcast counts down its arrival with near-certainty. Solar-flare radiation, by contrast, arrives in about 8 minutes.
Will a CME hit Earth, or miss it?
Each CME is drawn as a cone in its measured launch direction; Earth is hit only when it falls inside that cone. The catalog gives every CME a one-word verdict: Directed, Glancing, Misses, or Awaiting when its direction has not been measured yet. The 2012 near-miss replay shows a Carrington-class CME that crossed Earth's orbit and missed because Earth was elsewhere.
What is the difference between a solar flare and a CME?
A solar flare is a sudden burst of radiation (graded A/B/C/M/X by X-ray brightness) that reaches Earth in about 8 minutes. A CME is a slower, physical cloud of plasma that takes days to arrive. They often erupt together, but either can occur without the other.
I'm a radio operator. What does this mean for my bands?
Two separate hits. A flare ionizes the D-layer within minutes and absorbs HF on the sunlit hemisphere (NOAA's R scale) and passes within the hour. A CME arrival drives the geomagnetic storm: HF typically degrades for 24 to 48 hours (polar paths worst), while VHF operators may get auroral scatter on 6 m and 2 m. The radio conditions page rates 80 to 10 m day/night from the live SFI, sunspot number, A-index and Kp, shows blackout status, and warns about inbound CMEs days ahead. And every storm-capable CME's impact card and alert email carries a radio line.
What are the Kp index and the NOAA G-scale?
Kp is a 0 to 9 measure of global geomagnetic disturbance. The NOAA G-scale maps the peak Kp to storm severity: G1 (Kp 5, minor) through G5 (Kp 9, extreme). A higher Kp pushes auroras toward the equator, which is why CME Tracker sizes its aurora ovals by Kp. The header's Kp meter shows where the current value sits on the 0 to 9 scale; click it for a week of history and NOAA's 3-day forecast.
Is CME Tracker real-time, and where does the data come from?
Yes. CMEs and flares come from NASA DONKI, live solar wind and Kp from NOAA SWPC, and solar imagery from NASA SDO; historical replays use CDAWeb OMNI/STEREO and GFZ Potsdam. It is free to use, with no account required.
Can I ask an AI assistant like Claude or ChatGPT about solar storms?
Yes. CME Tracker can be connected to Claude, ChatGPT and other AI assistants. Once it is connected you can ask in plain words whether a solar storm is heading for Earth, when it should arrive, or whether you can see the aurora tonight from your town. The assistant answers from our live data and links back to the matching page here. It is free and needs no account. How to connect it.
For official space-weather forecasts, see NOAA SWPC.
Data and credits
Who the spacecraft are and what each one measures, with pictures: the spacecraft page. The feeds:
- CMEs, flares and shock arrivals: NASA DONKI.
- Solar images: NASA SDO, with archived frames from Helioviewer.
- Live solar wind, Kp, X-rays, protons, electrons, radio bursts and the aurora oval: NOAA SWPC. The L1 feed comes from SOLAR-1 since spring 2026, with ACE and IMAP as backups, and DSCOVR before that. SWPC also supplies the daily solar indices (10.7 cm flux, sunspot number, A-index) behind the header and the radio conditions page.
- Coronagraph images: NOAA's CCOR-1 on GOES-19 and CCOR-2 on SOLAR-1, through NOAA SWPC.
- Dst: the World Data Center for Geomagnetism, Kyoto.
- Kp history: GFZ Potsdam.
- Satellite orbits: CelesTrak.
- Cosmic rays: the Neutron Monitor Database.
- Historical solar wind: NASA CDAWeb (OMNI and STEREO).
Whether each feed is arriving on time is on the status page. What this site knows about you is very little, on purpose: see the privacy statement.
Put it on your own site: free embeddable widgets (plain iframes, no scripts) for the live map, the headline numbers, the aurora outlook and HF band conditions, plus an open JSON API. Pick one at Embed CME Tracker.
This is a one-person project and I'd genuinely like to hear from you: what's useful, what's confusing, a bug, or a feature you'd want. If you teach with it, please say so. I want to make it work better in the classroom.
Stay in the loop
Want to know when a storm is coming? Subscribers get a two-stage warning. First a heads-up days ahead when a significant Earth-directed CME is detected, stated as a range ("G1 likely, could reach G3"). Then, when the L1 monitor measures sustained storm conditions actually inbound, the arrival alert: about an hour before the storm reaches Earth, calling the band the measurement supports ("G1 to G2"). If the storm verifies at G2 or higher, you get one follow-up a couple of hours later. Misses are scored publicly on X and the site, not in your inbox. Plus the occasional feature note. No spam, unsubscribe in one click. (Privacy: we store your address and nothing else.)