Sólheimajökull, measured by name.

Minus ninety-three metres

That is the whole measurement. Minus ninety-three metres: the distance the western tongue of Sólheimajökull moved back during the glacier year 2024–25, recorded by Einar Gunnlaugsson, published as one row in one table in Jökull 75, the journal of the Iceland Glaciological Society. Ninety-three metres is a little short of a football pitch. It happened at one glacier, at one marked station, in one year, and a named person went out and measured it.

Almost nothing else written about this glacier is that precise. The most repeated figure — that it retreats about sixty metres a year — carries no station, no baseline year and no period, and matches no measurement we could find. It started on tour pages and has since reached Wikipedia.

This guide takes the opposite approach. Every chapter below is anchored to a figure with a date and an attribution: who measured it, when, and where it was published. That is unusually possible here, because Sólheimajökull is where Icelandic glacier monitoring began. In 1930 the meteorologist Jón Eyþórsson set markers at this snout and asked people nearby to read them each autumn. The network he started now reads about forty glacier termini regularly, and this one still reports.

The practical question is separate and more urgent. On 22 July 2026 the Icelandic Meteorological Office issued a hazard notice for the glacier margin: the ice ridge people had been using to reach the ice was coming apart. It had not been withdrawn when we last checked, on 30 September 2026. What it says, what happened around it, and what the booking pages say instead get their own chapters below.

Nobody from Hidden Corners has stood at Sólheimajökull. This guide is built from desk research: published measurement records, the Meteorological Office, peer-reviewed glaciology, a conference abstract, Icelandic reporting and operator pages, each dated where it matters.

What is established, and by whom

The glacier. An outlet of the Mýrdalsjökull ice cap, about eleven kilometres long and one to two kilometres wide, running from roughly 1,350 metres down to about a hundred, ending in a lake. Dimensions from Katla UNESCO Global Geopark.

The last measured year. −93 m at the western tongue, glacier year 2024–25, observer Einar Gunnlaugsson, in the annual front-variation report compiled by Hrafnhildur Hannesdóttir of the Meteorological Office for Jökull 75.

The long record. −977 m from 1930 to 1969, then plus 495 m from 1970 to 1995, then −750 m to 2010, leaving the snout 1,232 m behind its 1930 position. Schomacker and colleagues, Jökull 62, 2012.

The lake. About 0.45 km² by 2023, around sixty metres deep. Bathymetric survey work reported by Greta Wells and colleagues at the University of Iceland; the depth figure is the Geopark's.

The live hazard. Meteorological Office notice of 22 July 2026: unstable ice at the margin, rockfall from Jökulhaus above the access path. No update or withdrawal on the notice when rechecked on 30 September 2026.

Cost to stand there. No admission. Parking via the Parka app, ISK 750 or 1,000 per private car by class, plus an ISK 89 transaction fee.

Going onto the ice. Guided only, and in 2026 by a longer route than most booking pages describe.

The programme that started here

Iceland's glacier front-variation record is why this article can be written the way it is, and it began at this glacier.

In 1930 Jón Eyþórsson — later the driving force behind the Iceland Glaciological Society, founded in 1950 — set markers at Sólheimajökull and recruited people living near the ice to read them once a year. The method was a fixed point and a measured distance. The instruments have changed since; in 2024–25 the observer used two different distance meters to fix the snout.

What has not changed is the shape of the record: one station, one reading a year, one name against it. The measurements are made by Society members as volunteers, and for some years about forty termini have been read regularly; the 2024–25 report drew on returns from about forty stations.

The names matter to how the numbers read. In that same table, Öldufellsjökull is −75 m, measured by Jóhannes Gissurarson; Sléttjökull is −35 m, measured by Hrafnhildur Hannesdóttir herself; Kötlujökull, the outlet on the eastern side of the same ice cap, is −30 m, measured by Jóhannes Marteinn Jóhannesson. Four outlets of one ice cap, four figures, four people, one year. A glacier does not have a retreat rate. A station has a reading.

The report's summary line is meirihluti jökulsporða hörfa: the majority of termini are retreating, most of all on the larger outlet glaciers. The majority, not all. That qualification is the programme being honest about its own data.

Back 977 metres, forward 495, back 750

The long record comes from one paper. Reconstructing the marginal zone from the Society's series, aerial photography and field mapping, Anders Schomacker and colleagues published the glacier's twentieth century in Jökull 62 in 2012. It moves in three phases, and only two of them are retreat.

1930 to 1969: minus 977 metres

A mean of about twenty-five metres a year across four decades — well under half the figure the tour pages quote as an average.

Two things that 977 metres is not. It is not the glacier's recent maximum: Schomacker's team places the Little Ice Age margin roughly two kilometres beyond where the ice stood in 2010, so the 1930 baseline was already a retreated position. The record begins in the middle of the story, as instrumental records generally do. Nor is it a steady process. It is a total of individual years the Society logged separately, some large and some near zero, and keeping those annual readings is the only reason anyone can see what came next.

1970 to 1995: plus 495 metres

For a quarter of a century the ice came back down the valley by half a kilometre. It is the most useful single fact here for anyone trying to judge a glacier statistic. The paper ties it to a period of relatively cooler climate between about 1970 and 1995.

The snout did not only lengthen, it fattened: 70 to 100 metres of thickening at the margin between 1960 and 1996. The glacier a visitor saw in the mid-1990s was longer and considerably deeper than the one measured in the 1960s.

So "retreating since 1930" is a simplification the instruments do not support — and an inconvenient advance is evidence that the record is recording the glacier rather than the expectation. A programme that only ever produces the number people want is not a monitoring programme.

Watch what happens to these figures downstream. The Geopark's page dates the advance to 1969–1995 and puts the 2010 margin about a kilometre behind 1930, where the paper says 1,232 metres. Neither is a serious error. Both show how fast a measured figure loosens once it leaves the paper it came from.

1995 to 2010: minus 750 metres, and a thinner snout

About fifty metres a year, roughly double the 1930–1969 rate. The vertical figures matter more. Having thickened by 70 to 100 metres, the snout thinned by 120 to 150 metres between 1996 and 2010: fourteen years undid thirty-six. By 2010 the margin was 20 to 40 metres thinner than in 1960 and 1,232 metres behind where the markers had found it in 1930.

Thinning is the change a visitor can actually see, which is why the valley walls deserve more attention than the ice. Where rock is scoured pale below a line and weathered above it, that line is the former ice surface. A hundred and fifty metres is a long way to look up.

It also set up everything since. A snout that loses that much thickness stops resting on the valley floor the way it did — and the floor beneath this one turns out to be a deep trough.

Three stations, one glacier, one observer

Ask what Sólheimajökull did in a given year and the answer depends on where the question is pointed — and for a long stretch of the record there were three places to point it.

In Oddur Sigurðsson's front-variation report for 2003–04, published in Jökull 55 in 2005, Sólheimajökull appears as three rows: vesturtunga, the western tongue, at −87 metres; Jökulhaus, at the rock mass in the middle, at −58; and austurtunga, the eastern tongue, at −39. Same glacier, same season, a spread of forty-eight metres.

Two details make that table more instructive. The dates: readings on 21 October 2003 and 22 October 2004, a year and a day apart, which is the discipline that makes an annual figure mean anything. And the observer: all three stations were read by the same person, Einar Gunnlaugsson of Reykjavík. The spread is not three people disagreeing. It is one person, one method, the same two days, and a snout that is not a straight edge. The report adds that no two consecutive years had been as hard on this snout since 1930, with close to 200 metres lost across the pair.

By 2024–25 only the western tongue appears as a data row, though the report notes that Einar and Gunnlaugur, father and son, climb Jökulhaus every year for overview photographs. The consequence is direct: the −93 metres in the current report and the −87 from twenty years earlier are the same station, so they are comparable. Compare either with a figure that names no station and you are not comparing anything.

And the observer's name is the same in 2025 as in 2005. Twenty-two years of one glacier, read by one person, published where anyone can check it.

The class that measures in October

There is a second, entirely separate series here, and it is kept by twelve-year-olds.

Since 2010, seventh-years from Hvolsskóli, the school at Hvolsvöllur, have gone to Sólheimajökull each October and measured from the interpretive sign to the edge of the ice. In 2010 it was 318 metres. In 2021 it was 726. That is 408 metres of retreat in eleven years, recorded with a tape and a fixed reference point — Jón Eyþórsson's method, unchanged.

The annual values inside that total are wildly uneven. On 8 October 2018 the class measured 110 metres in a single year, which local reporting linked to heavy calving in the preceding months. In the year to October 2021 they measured eleven. A decade average of about forty metres a year is arithmetically true and describes no year that happened.

In 2018 the municipality named the teacher behind the project as Jón Stefánsson, with the Dagrenning rescue team taking pupils onto the lake to measure its depth. By 2021, as the head teacher Birna Sigurðardóttir described it, the class that once measured on dry land needed a boat.

The two series do not agree, and we will not pick between them. Covering the 2019 trip, Agence France-Presse reported the Society's figure for the preceding year as around 200 metres, roughly double what the class had measured in the same season. Different definitions of the margin, different points measured from, and a snout with two tongues will all do that. We found no published reconciliation.

What the children's record has that the professional one does not is that the measurement is a walk: a distance covered on foot, from a sign that has not moved, to ice that has.

Where "sixty metres a year" comes from

Now set the claimed figures beside the measured ones. The same exercise works on every glacier statistic in Iceland.

The claim, still live on a large operator's attraction page when we rechecked it on 30 September 2026, is that the glacier is "retreating at an average rate of about 60 meters (200 feet) annually." No station, no baseline year, no period.

The measured figures, all cited above: about 25 metres a year as the 1930–1969 mean; a negative rate from 1970 to 1995, because the glacier advanced; about 50 metres a year for 1995–2010; −87 metres at the western tongue in 2003–04 and −93 at the same station in 2024–25; about 40 metres a year as the schoolchildren's decade average, containing years of 110 and 11.

There is no year in which sixty metres is the measured value and no defensible period over which it is the mean. What lifts this above pedantry is that the same page carries its own contradiction: it also states the glacier "retreated 592 m (1,942 ft) since 2010." Over the fifteen years to 2025 that is a mean of about thirty-nine metres a year. The page's two numbers disagree with each other by more than half again, and nobody appears to have noticed.

The figure travels. Wikipedia's Sólheimajökull article says the glacier "is melting rapidly around 60 metres (200 ft) per year," and the citation offered is an Icelandic tour company's website. A marketing round number has been laundered into a reference work, where the next writer will find it.

Other published figures slip in a different way. In October 2018 the municipality's own report of the school measurement said the glacier had retreated about 697 metres since 2010. On the class's series that cannot be the retreat — the ice stood 318 metres from the sign when they started — so 697 is almost certainly the distance from the sign that autumn.

That is the real failure mode, and it is not dishonesty. "How far the ice is from the sign" and "how far the ice has moved" are different measurements that produce similar-looking numbers, and one sounds more alarming.

One sentence to carry away: over roughly the last century the snout has moved back on the order of one and a half to two kilometres, unevenly, with a twenty-five-year advance in the middle, and recent annual figures at the western tongue are in the tens of metres. Less quotable, considerably more true.

The lake that was not there

Every figure so far has been horizontal. The most consequential measurement here now is of something that is not ice at all.

A proglacial lake occupies ground the snout has vacated. It did not exist within living memory. Greta Wells and colleagues at the University of Iceland's Institute of Earth Sciences date its formation to around 2007; Katla Geopark says 2011. Both are published and attributable, and they are four years apart. We report the conflict rather than pick a side.

What the lake has done since is measured rather than estimated. Wells, with Þorsteinn Sæmundsson, Eyjólfur Magnússon, Guðfinna Aðalgeirsdóttir and Finnur Pálsson, reported field surveys of lake bathymetry and subglacial topography covering volume evolution from 2009 to 2023. By 2023 the lake covered about 0.45 square kilometres. The Geopark gives a present depth of around sixty metres; one operator page says sixty to seventy.

Two decades is not long for a body of water half a square kilometre in area and sixty metres deep to appear in a valley; the far larger lagoon at Jökulsárlón began forming in the 1930s.

The lake also changed what the schoolchildren's measurement physically is. The class that walked to the ice on dry land in 2010 now needs a boat, which says more about the numbers here than most of the writing about them.

What the model does with the lake

Here the article crosses a line most writing about glaciers leaves unmarked. Everything above was measured. Everything here is modelled, and the difference is not a technicality.

The reason the Wells group surveyed the bed and not only the water is that the shape of the rock beneath the ice determines what the lake becomes, and what they found is an overdeepened trough. Feeding continued retreat into that geometry, the model says the lake will roughly double its maximum depth, quadruple its surface area and increase its volume ninefold, reaching a maximum extent about four kilometres up-valley from where the water ends today.

Those are conditional figures: what the basin can hold, not a date on which it will hold it. The timing is framed separately: if recent retreat rates continue, the terminus could enter the deepest part of the trough in approximately a decade and retreat out of the lake basin within a century, and the authors add that it could happen faster if calving increases in deeper water.

That last clause is why the physics has changed. A glacier ending on gravel retreats when it melts. One ending in deep water also calves: blocks break off, float away and are gone, and the loss becomes partly a function of water depth rather than of how warm a summer was. Once the terminus is in the deepest trough, the valley floor does more of the work than the weather.

The honest statement is therefore layered. The lake's present size is a survey result. Its potential size is a model output conditioned on bed geometry. The timescale is a projection conditioned on retreat rates continuing, which the record above shows they have not reliably done. Collapse those three into one confident sentence and you have discarded the distinction the work was done to establish. It happens: the Geopark's page says the lagoon could become up to four kilometres long "in the next few decades", where the abstract's own timescale runs to a century.

The hazard the lake invented

The same modelling identifies a consequence that matters more to a visitor than the retreat does, and it did not exist here before the water did.

As the lake extends up-valley it will run beneath several stretches of valley wall the survey identifies as having high topographic potential to source a rockfall or an avalanche. A large enough mass falling into deep water displaces it, and the displacement wave can become a glacial lake outburst flood — a flood with no volcano behind it, from a lake less than twenty years old.

Iceland has a documented precedent for the mechanism at another glacier. At Svínafellsheiði in the south-east, Stephanie Matti and Helga Ögmundardóttir described a fracture roughly 1.7 kilometres long with some 60 million cubic metres of rock in motion above Svínafellsjökull, cliffs 400 metres above the ice, and a modelled wave in the proglacial lake of about 75 metres. Their point, published in 2021, was that Iceland has no record of this type of hazard affecting people or infrastructure — which is exactly why it is easy to leave out of a risk assessment.

Nothing on that scale is documented above Sólheimajökull and nobody has claimed it. What the work here establishes is narrower and still significant: the receiving basin is being created now, and the slopes it will reach have been identified. This is a hazard with a lead time, which is an unusually useful thing to have.

The notice of 22 July 2026, in the present tense

A hazard notice is in force at the margin of Sólheimajökull. The Icelandic Meteorological Office published it on 22 July 2026. When we rechecked it on 30 September 2026 it carried no update or withdrawal, and we found no superseding notice.

What it says, as precisely as we can render it. A small lagoon has formed east of Sólheimajökull's tongue. Together with the main proglacial lake, that new lagoon delineates an ice ridge forming the connection from the shore to the glacier. Traversing this ice ridge "now poses significant danger." There is a risk of unstable ice formations giving way, and a risk of slipping and falling into the cold, deep glacial lagoons on either side.

Separately, the Office draws attention to the risk of rockfall from Jökulhaus, the mountain overlooking the access route on the south-eastern side of the glacier. Rockfall can occur along most of the length of the hiking trail. It happens suddenly, cannot be accurately predicted, and the hazard rises with heavy precipitation and freeze-thaw.

The notice's framing of who is at risk is unusually direct for an official document: "inexperienced travellers can easily find themselves in hazardous situations in the vicinity of retreating outlet glaciers." It also records what the industry was doing at the time — local mountain guides monitoring the changes, and tour operators having "currently cancelled trips as a precaution while looking into new and safer routes."

Read against the preceding chapters, this is the modelled hazard arriving in ordinary language. An ice ridge with deep water on both sides is the lake's new eastern arm meeting the old one. An unpredictable slope above a path is the rockfall potential the survey mapped, at a scale that affects an individual walker rather than a valley.

Eighteen days, and the pages that did not change

The notice did not arrive out of nowhere, and the response to it is more interesting than the notice.

12 July 2026. Ten days before the Office published anything, Stephan Mantler — chairman of Fagráð, the professional safety council set up for glacier hikes and ice-cave tours in Vatnajökull National Park in 2024, and originator of the GLACIS glacier risk-assessment platform — published a technical assessment of the Sólheimajökull access route. He described the crossing as depending on a layered and possibly unstable ice structure bounded by open water, and expected it to become impracticable within weeks, "if not days," through undercutting and tunnel formation near the choke point. He also noted sinkholes and open water in the dead ice along the trail, and that it was unclear whether the bridge was grounded or floating.

21 July 2026. The Icelandic Tourist Board convened authorities and operators. The meeting ran under the Chatham House Rule, which is why no individual is named in the reporting. Per Iceland Review's account of 27 July, it agreed not to close the site: temporary trail or area closures were discussed and judged "not warranted at this stage." The agreed response was clearer multilingual signage, messaging that glacier travel requires certified guides, improved walking routes and trail infrastructure, possible engineering solutions, rescue equipment near the lagoon, streamlined permits for the works, and co-ordination of signage by SafeTravel. Guided tours were to continue on daily professional risk assessments.

22 July 2026. The Meteorological Office notice, above.

23 July 2026. Visit South Iceland published its own notice, materially more restrictive than the meeting's conclusion: visitors "are strongly advised to avoid the area altogether." Two official-sounding sources, two days apart, pointing different ways.

Around 23 July 2026. A new approach trail opened, running through a valley that circles south and east around Jökulhaus before descending towards the glacier. Writing on 30 July, Mantler records that it adds about an hour to the approach and that a short exposed section near the top has been fitted with expansion bolts and eyelets for a fixed hand line. He also records that the ice bridge degraded significantly between 8 and 29 July, that meetings had run over three weeks between operators, landowners, the Tourist Board, the Meteorological Office and search and rescue, and that a zodiac already on the lake for kayak tours could double as a rescue craft.

Since then. Verified customer reviews on a large booking platform, dated 30 July, 4 August and 3 September 2026, describe completed Sólheimajökull glacier walks with crampons, axes, ropes and harnesses. The 3 September reviewer found the hike "a bit involved" and not for anyone afraid of heights — consistent with the longer route.

And the booking pages. Through late September 2026, operator pages were selling Sólheimajökull glacier hikes and ice climbing as year-round products with no visible reference to the hazard notice, the route change or the extra hour. One attraction page still describes reaching the viewpoint as "a 15-minute (500-meter) walk from the parking lot." That fits the old route, not the one Mantler describes.

None of that is evidence of an operator behaving badly, and we name none of them. Walks are demonstrably running, the authorities explicitly decided against closure, and suspensions were always company-by-company and day-by-day. The point is narrower and it is the practical core of this guide: a bookable product is not a statement about today's conditions, and a page describing a fifteen-minute walk is not a statement that the walk is still fifteen minutes. The Meteorological Office notice and SafeTravel are statements about conditions, both are free, and one has been live since July.

So: check vedur.is and SafeTravel before you travel and again the day before, and ask your operator when their most recent group reached the ice, by which route, and how long the approach now takes.

What Katla is measured for

Mýrdalsjökull sits on Katla, and Katla is measured continuously — but measured is not predicted, which is the same distinction this article has been making throughout.

The established figures, from the Meteorological Office: an ice-filled caldera about nine by fourteen kilometres; at least twenty-one eruptions in eleven hundred years; an average repose period of about fifty years; a last eruption that broke the ice surface in 1918. That is 108 years ago, more than twice the average interval. It does not follow that an eruption is due. An average repose period describes a past distribution, not a countdown, and the Office does not present it as one.

What actually reaches visitors is the jökulhlaup, the glacial outburst flood, which happens without an eruption. Geothermal cauldrons melt ice from beneath, meltwater pools, the seal fails. The monitoring that matters is therefore hydrological and chemical as much as seismic: river conductivity, water level, gas.

• 18 July 1999. A flood down Jökulsá á Sólheimasandi, the river this glacier feeds, preceded by five hours of tremor and by fifteen cauldrons appearing on the ice. Modelled peak discharge near 4,780 cubic metres a second at the terminus; 1,700 recorded at the bridge four kilometres downstream an hour after onset; back near base level within about six hours. It moved some 470,000 cubic metres of sediment and left blocks over ten metres across.
• 9 July 2011. Four cauldrons drained into Múlakvísl on the far side of the ice cap and took out the Route 1 bridge. No eruption.
• 8–11 July 2025. Cauldrons 13 and 14 drained into Leirá Syðri and Skálm, with raised conductivity, hydrogen sulphide and a gas pollution warning. No eruption.
• 18 August 2026. A jökulhlaup began in Múlakvísl. Conductivity rose from about 285 microsiemens per centimetre on 19 August to just over 500 on 20 August, a level last reached in the July 2017 flood. The Láguhvolar monitor, roughly two kilometres from the margin of Kötlujökull, recorded hydrogen sulphide above the health-protection limit overnight. By the evening of 20 August the Office judged the flood to be subsiding, but on 1 September it reported conductivity still high for the season and hydrogen sulphide again above the limit. No eruption.

The 2026 event is the correction to a tidy pattern: three of these are July, that one is August. Clustering in high summer is real and worth knowing. "Always July" would be exactly the kind of round claim this article exists to argue against.

What to do about it is cheap. Almannavarnir's Katla response plan models flood routes onto Sólheimasandur, Mýrdalssandur and west towards Markarfljót, and states that in many places evacuation time is short. Alerting is by SMS to phones in the area, so keep yours charged, connected and switched on while parked on a sandur. A sulphurous smell near a glacial river is a measurement your nose is taking: leave the bank and move to higher ground. Gas collects in hollows on calm days.

The part that is not measured: the rules

About 2.27 million foreign visitors left Iceland through Keflavík in 2025, and the Tourist Board's survey put roughly four in five of them in the south at some point. Sólheimajökull takes a share of that because it is the easiest ice in the country to reach.

The statutory scaffolding underneath is thinner than a booking page suggests, and unlike everything else in this article it is not quantified anywhere.

Under the Act on the Icelandic Tourist Board, No. 96/2018, an operator running organised tours must hold a safety plan for each tour type: risk assessment, work procedures, contingency plan, incident reporting. The Board can audit and can levy daily fines of ISK 50,000 to 500,000. But the plans are not routinely submitted and are not normally reviewed for adequacy; the enforcement trigger is a plan found clearly inadequate, usually after the fact.

There is no binding legal standard for who may lead people onto a glacier. The Tourist Board's requirements on guide knowledge, experience and training for glacier walking are leiðbeinandi reglur — advisory — issued after a safety bill failed in parliament in 2012. Certification is handled by the industry's own association. Vakinn, the Tourist Board's quality scheme, audits operators who choose to join it. Choose is the operative word.

The gap became visible on 25 August 2024, when an ice wall collapsed during a summer ice-cave tour at Breiðamerkurjökull, killing an American visitor and critically injuring his partner. Vatnajökull National Park responded by imposing conditions on operator contracts from autumn 2024: a designated lead guide, daily joint safety assessments of each cave and of overhanging ice, photographic documentation, and a professional oversight council funded by per-client fees. That council is Fagráð.

Here is the structural fact that makes Sólheimajökull different from the glaciers most coverage discusses. Those measures are contractual conditions of access to a national park, and Sólheimajökull is not in one. In April 2025 a safety specialist told Iceland Review that there are no established rules in Iceland for how ice-cave tours should be conducted outside Vatnajökull, that the Tourist Board lacks authority to make them, and that a ministerial regulation or an act of parliament would be required. We found nothing indicating that changed during 2026.

So in July 2026 the chairman of the park's safety council was publishing hazard assessments of a glacier where his council has no jurisdiction, and operators rerouted an access trail through voluntary meetings with landowners and agencies. It worked. It worked because competent people chose to make it work, not because a rule required it — and that difference matters when you are choosing who to walk with.

This is not an argument against guided glacier walking, which is the only safe way onto ice and is done well by many Icelandic operators. It is an argument for asking: who specifically is leading, what certification they hold and from whom, what the daily assessment process is at this glacier, when the last group reached the ice and by which route, and what conditions cause a cancellation.

What Sólheimajökull is not

A good deal of confusion about this glacier is really confusion between four Icelandic things with overlapping names. Worth clearing up before you plan anything.

It is not Vatnajökull, and its caves are not the blue crystal caves. The famous winter ice caves are in Vatnajökull, Europe's largest ice cap, several hours further east and inside a national park with a different regulatory regime. If that is what you came for, the Vatnajökull caves are their own subject. Meltwater caves do form at Sólheimajökull in winter, in a different place each year, and they are guided; they are not the same phenomenon and not reliably there.

It is not the Katla ice cave. The cave marketed under that name is in Kötlujökull, the outlet on the eastern side of the same Mýrdalsjökull ice cap, reached from Vík rather than from this car park. Same volcano underneath, different tongue and a different tour.

It is not in a World Heritage site. Iceland's glacier World Heritage property is Vatnajökull National Park, inscribed in 2019. Sólheimajökull sits in Katla UNESCO Global Geopark, a different UNESCO programme with a different purpose. This is where "gone by 2050" comes from: a 2022 UNESCO assessment found glaciers at about a third of World Heritage sites will disappear by 2050. That is a statement about specified properties, and this glacier is not one.

It is not a volcano you can watch. Katla is under the ice cap, invisible from the car park, and has not broken the ice surface since 1918.

What it is: the most accessible glacier snout in Iceland, the oldest continuously monitored one, and currently the one with a live hazard notice on its access route.

Okjökull, and the two-hundred-year number

The last category is projection, and the most quoted projection about Icelandic glaciers is engraved on a rock about 150 kilometres from here.

Okjökull, on Ok in the west, covered about 16 square kilometres in 1890 and 0.7 by 2012. In 2014 Oddur Sigurðsson of the Meteorological Office — the same person who compiled the 2003–04 front-variation report quoted above — struck it off the register: no longer a living glacier, only dead ice, not moving. In August 2019 a plaque was set into the bare rock, addressed to the future, with the May 2019 atmospheric carbon dioxide figure of 415 parts per million stamped at the bottom. The prime minister, the environment minister and a former UN High Commissioner for Human Rights walked up for it.

The sentence usually cropped from the coverage is the operative one. The plaque does not say the glaciers are going in a decade. It says that in the next 200 years all of Iceland's glaciers are expected to follow the same path. The most-quoted monument to glacier loss in the world is more conservative than most of the journalism quoting it.

The peer-reviewed projections are more conservative still and disagree with each other, which is what projections at this range do. Compagno and colleagues put the loss of Icelandic glacier volume by 2100 at 43 ± 11 per cent under a low-emissions pathway and 85 ± 7 per cent under a high one — substantial, and not disappearance. Vatnajökull National Park's material describes roughly half of Iceland's glacier volume gone by 2100, with only small high ice caps after 200 years. Reviewing the modelling in 2020, Guðfinna Aðalgeirsdóttir and colleagues cite projections to 2100 of about 80 per cent for Langjökull, 60 for Hofsjökull and 20 to 30 for Vatnajökull, while noting these probably understate the loss. Katla Geopark puts the complete disappearance of Sólheimajökull at 100 to 200 years.

None of those is 2030 and none is 2050. The spread between them is the honest state of the science, and it is wider than any headline you will read.

What to read from the shore

Assuming the shore is as far as you get — which for many visitors this season will be the sensible maximum — there is more to read here than at any comparable roadside glacier, because you can see the measurements and not just the ice.

The distance to the ice. From the interpretive sign to the margin is the retreat made physical. The schoolchildren's marker moved from 318 metres from that sign in 2010 to 726 in 2021. Pace some of it out; the figure means more in your legs than in your notebook.

The trim line. On the valley walls, find the boundary between bare scoured rock and weathered or vegetated ground above. That marks the former ice surface. You are looking up at the 120 to 150 metres of thinning measured between 1996 and 2010.

The black bands. The dark layers in the ice are tephra — ash from Katla and its neighbours, each band a deposit from one eruption, buried by later snow and carried down-glacier. A stratigraphy visible from the ground, and the reason this glacier looks striped rather than white.

The lagoon's edges. Look for undercutting where water meets ice, and for blocks floating clear. Both are calving, the process that now drives the retreat more than the weather does.

The slope above the path. Jökulhaus is the rockfall source named in the 2026 notice. Fresh rock on the ground — unweathered, no lichen — is recent. Do not linger below the steepest sections.

What not to do. Do not step onto ice, and specifically not onto any shoreline ice ridge, without a guide. Do not approach the lagoon edge: the ground is undercut and the water is around sixty metres deep and near freezing. Do not treat other people being out there as information about whether it is safe.

When to go

Season changes what you can do here more than what you can see, and it interacts directly with the hazards above.

Summer, June to August. Long daylight, easiest driving, fullest tour operation in a normal year. Also fastest melt and the least stable margin — the 2026 notice is a July document — and the months when Katla's meltwater outbursts have historically occurred, including the August 2026 event in Múlakvísl. Busiest by a wide margin.

Winter, roughly November to March. When meltwater ice caves exist here, where conditions and operators permit; they form somewhere different each year and are never guaranteed. Daylight is very short, the approach icier, the road weather-affected. Operators supply crampons for the ice. Nothing supplies grip for the drive.

Spring and autumn. The quieter middle. September and October give reasonable daylight and thinner crowds, and — fittingly — October is when both measurement programmes take their annual readings.

Any season. South Iceland weather governs, and wind changes more plans here than temperature. Check vedur.is for forecasts and hazard notices and road.is for conditions, and treat a strong wind warning as a reason to move the day rather than a challenge.

For 2026–27 specifically. Plan a viewpoint visit and treat a walk on the ice as a bonus you confirm on the day. The approach is longer than most published descriptions say, and the hazard notice has not been lifted.

Location and access

Driving. Route 1 east from Reykjavík, roughly two hours, then the signposted turn onto Road 221 and about four kilometres to the car park at the road end. The side road is unpaved but ordinarily passable in a normal car.

Parking. Charged through the Parka app at parka.is, rechecked 21 September 2026 and unchanged. Six classes, assigned by the vehicle's registered seating capacity: ISK 750 (class A) and ISK 1,000 (class B) for private vehicles, ISK 1,800, 3,500 and 6,400 for buses by size, ISK 300 for motorcycles, plus an ISK 89 transaction fee. You may pay up to 24 hours after the parking period ends; an unpaid charge attracts an ISK 1,960 late fee and a claim against the registered owner. No admission charge for the site.

Iceland's parking-fee industry at natural attractions is contested. In June 2025 Neytendastofa, the Consumer Agency, fined Parka Lausnir ehf. ISK 1,000,000 over price and boundary information at its sites; Sólheimajökull was not among the locations found in breach. Photograph the sign and your plate, and pay inside the window.

The walk in. Published figures disagree, and this is not a quibble. One large operator's attraction page describes a 15-minute, 500-metre walk to the viewpoint; tour descriptions of the walk to the ice run longer; and the approach to the ice was rerouted in late July 2026 around the south and east of Jökulhaus, adding about an hour, with a short exposed section near the top fitted with a fixed hand line. Assume the longer figures until an operator tells you otherwise.

Without a car. Day tours from Reykjavík combine the glacier with other south-coast stops, and this is how most visitors arrive. In a period of uncertain access it has a real advantage: the operator absorbs the decision about whether the ice is reachable, and carries the liability for it.

The map on this page shows the car park and glacier for orientation only. It shows no route onto the ice, and no pin establishes that any path or ice ridge is currently safe.

Budget: one published price, and two decisions

Exactly one cost here has a current published source. Everything else is a choice.

• Standing at the viewpoint: parking only. ISK 750 or 1,000 per private car by class, plus the ISK 89 transaction fee, per the Parka page rechecked on 21 September 2026. No admission.
• A guided glacier walk: a per-person price from a licensed operator, typically a few hours with crampons, harness, helmet and axe supplied. We quote no figure: prices vary by operator, season and group size, and we hold no verified current quote. Compare on guide-to-client ratio, time actually on ice as against total duration, whether stiff boots are included or hired, and the cancellation policy — which matters unusually much this season.
• A day tour from Reykjavík: a bundled coach price covering transport and several south-coast stops, sometimes with the glacier walk as an add-on. Cheaper than car hire for one traveller; more expensive for four.

The variables that move your total are car hire and whether you go on the ice at all. One further 2026 consideration: the rerouted approach adds roughly an hour, which lengthens tours and eats into a south-coast day. Budget the time as well as the money.

The calculator on this page covers the parking charge only, because that is the single figure with a current published source, and it is per vehicle rather than per person. It is not a trip budget.

Booking: what to ask before you pay

Because the statutory standard for glacier guiding is advisory rather than binding, and because the access route was rebuilt mid-season in 2026, the questions matter more here than the price comparison.

• When did your most recent group actually reach the ice, and by which route?
• How long is the approach now, and does your published duration reflect the route change?
• Who will lead the group, what certification do they hold, and from which body?
• What is your daily safety-assessment process at this glacier, and what conditions cause you to cancel?
• What is the guide-to-client ratio, and how much of the stated duration is on ice rather than walking in?
• Are stiff-soled boots included or hired separately?
• If conditions stop the walk, do I get a refund, a credit or a substitute?

A bookable product is not a statement about today's conditions. The Meteorological Office and SafeTravel are, and both are free.

We carry no affiliate links or commercial relationships for this destination, and no operator is named or recommended anywhere in this guide.

Packing for the glacier margin

Two kits, depending on whether you are going onto the ice, and one constant: south Iceland weather arrives sideways.

For the viewpoint, any season. A genuinely waterproof shell with a hood and a warm mid-layer. Waterproof trousers, because horizontal rain makes a mockery of a jacket alone. Boots with grip and ankle support for loose glacial debris — trainers are the commonest mistake here. Hat and gloves; wind chill off ice is significant even in July. A charged, connected phone, because civil-protection alerting is by SMS. A payment card and your registration number for the parking app.

Additionally for a guided walk. Operators supply crampons, harness, helmet and axe, and usually require stiff-soled boots, which most also hire out; confirm in advance rather than arriving in soft trainers and being turned away. Layers you can shed, because walking on ice in waterproofs is warm work. Sunglasses in bright conditions; glare off ice is real. Gloves you can still work a buckle in. Allow for a longer approach than older tour descriptions suggest.

Winter additions. Micro-spikes for the walk in from the car park — not a substitute for the crampons the guide provides on the ice. A headtorch, since daylight is short. Extra insulation for standing still while a group is fitted with equipment.

What no kit does. None of this makes an unstable ice ridge safe to cross, and none substitutes for a guide. Crampons at an undercut lagoon edge are worse than useless: they get you further out.

Guided operators supply and fit technical on-ice equipment; confirm what your booking includes rather than buying it for one visit.

A considered half-day

Sólheimajökull rewards being treated as a deliberate stop rather than a tick on a south-coast dash, and the difference is about an hour.

Leave Reykjavík early enough to arrive before the mid-morning coaches. From the car park, walk out slowly and spend the first ten minutes on the valley walls rather than the ice: find the trim line, work out how much higher the surface used to be. Then the lagoon edge, then the tephra bands. If you can find the interpretive sign, the distance from it to the margin is the measurement this whole guide is about, and pacing part of it out is the best thing you can do here.

If you are doing a guided walk, budget well beyond the older published durations: the rerouted approach adds about an hour each way, the group moves slowly, and the guide will stop to check ground. If you are not, an hour at the margin covers everything above and is more than most visitors give it.

Combining. The glacier sits between Skógafoss to the west and Vík to the east, and the black-sand coast at Reynisfjara is half an hour on — a beach with its own serious and frequently underestimated hazards, which our guide to Reynisfjara sets out. Doing both in a day is normal. Doing both plus two waterfalls plus a glacier walk is the itinerary that produces people running back to a coach.

If the ice is not reachable and you want a second measured landscape rather than a second waterfall, Katla Geopark's other geosites along this coast are the better substitute.

Further reading

Chosen from publishers' descriptions and bibliographic records.

• Andri Snær Magnason, On Time and Water, translated by Lytton Smith (Serpent's Tail, 2021 in the UK). An Icelandic writer on glaciers, deep time and the difficulty of grasping change on a scale larger than a lifetime. It is the text behind the Okjökull plaque and the best single preparation for standing here. Publisher page.
• Jökull, journal of the Iceland Glaciological Society. Not a book but the actual source: the annual front-variation reports are open access, in Icelandic with English summaries and readable tables. Every station figure in this article can be checked there in a few minutes, with the observer's name against it. jokull.jorfi.is.
• Þorleifur Einarsson, The Geology of Iceland: Rocks and Landscape, translated by George Douglas (Mál og menning, ISBN 9789979306894). The standard English-language introduction to Icelandic geology, covering glaciers, volcanism and the interaction that produces jökulhlaups. We found no live publisher page; the library catalogue record confirms the title and translator.

Free, current, and more useful than any book for an actual trip: the Icelandic Meteorological Office for hazard notices and weather, SafeTravel, run by the Icelandic Association for Search and Rescue, for conditions and alerts, and road.is for the drive.

Questions travellers ask

Can I visit Sólheimajökull right now?

The site is not closed. Authorities and operators agreed on 21 July 2026 that closure was not warranted, choosing signage, route improvement and daily professional risk assessments instead. A Meteorological Office hazard notice of 22 July 2026 remained live and unwithdrawn at our last check on 30 September 2026, covering unstable ice at the margin and rockfall from Jökulhaus. Guided walks have run since, on a rerouted approach: verified reviews dated 30 July, 4 August and 3 September 2026 describe completed tours.

How long is the walk from the car park?

Published figures disagree. One attraction page says fifteen minutes and 500 metres to the viewpoint; and the approach to the ice was rerouted in late July 2026 around Jökulhaus, adding about an hour and including a short exposed section with a fixed hand line. Assume longer, and confirm with your operator.

Can I walk on the glacier without a guide?

No. Crevasses, an unstable margin and deep cold water make this unsurvivable ground for the unequipped, and the 2026 notice concerns precisely the approach unguided visitors had been using.

How fast is it retreating?

It depends which station and which period. The most recent published figure is −93 metres at the western tongue for the glacier year 2024–25, measured by Einar Gunnlaugsson for the Iceland Glaciological Society. The long record gives about 25 m/yr for 1930–1969, an advance from 1970 to 1995, and about 50 m/yr for 1995–2010. The widely quoted "about 60 metres a year" matches no measured year or period we could find.

Will it be gone by 2030 or 2050?

No projection we could find says either for this glacier. Katla Geopark puts complete disappearance at 100 to 200 years. The 2050 figure appears to come from a 2022 UNESCO assessment of World Heritage sites, and Sólheimajökull is not in Iceland's glacier World Heritage site.

Is this where the blue ice caves are?

No. Those are in Vatnajökull, hours further east, inside a national park with a different safety regime. Meltwater caves do form here in winter, somewhere different each year, and they are always guided.

Should I worry about Katla?

Not about an eruption, which is unpredictable and not imminent on any published assessment. The relevant hazard is a meltwater outburst flood, which happens without one — most recently in Múlakvísl from 18 August 2026. Keep your phone on for SMS alerting, and leave a riverbank if you smell sulphur.

What gets measured next, and by whom

If the school's tradition holds, some time this October a class of seventh-years from Hvolsskóli will drive east, walk to the water and find out how far the ice has gone since last autumn. They will need the boat.

Some time in the same season Einar Gunnlaugsson will take a reading at the western tongue, as he has since at least 2003, and the figure should appear in the next front-variation report with his name beside it and about forty other stations around it. The Meteorological Office's conductivity sensors on Múlakvísl and its gas monitor at Láguhvolar will keep reporting whether or not anything happens. If the Wells group returns to the lake, the bathymetry will be resurveyed against the 2009–2023 baseline, and the model of what the trough can hold will be tested against what it does.

None of that is spectacular and none of it produces a round number. It produces rows: a station, a date, a value, a name. That is what this glacier offers that the photographs do not — not a feeling about loss, but a distance somebody walked and wrote down, published where you can check it against whatever you were told.

The ice will be somewhere in October. Somebody will measure where.

Dated photographs show the place, details and wider setting. They are not evidence of present access or conditions.