Forecasting La Palma’s 2021 eruption after the fact
Opinion piece
Written by Ben Ireland
PhD student in Volcanology, Bristol University and VolcanoStories Editor, GeoTenerife
Share on:
In the years since the 2021 Tajogaite eruption on La Palma, volcanologists have published research identifying precursory signals for both the beginning and end of the eruption from a variety of different data sources. One recent study has even shown evidence that the reactivation of La Palma’s magmatic system occurred 10-15 years before the eruption. This has led some local commenters to question why these insights were not available and shared with the public before the eruption, and question if data has been withheld from the public. The reality is that the lack of these insights until after the eruption is because many methods used in volcanology rely on measurements that can only be taken after an eruption has begun.
Let’s explore the ‘hindcasts’ that have been made for various aspects of the eruption, and how these may be useful for future eruptions on La Palma and beyond.
The lack of forecasting by volcanologists of various aspects of the 2021 Tajogaite eruption on La Palma, and the knock-on impacts on the population, has been the source of intense scrutiny towards scientists and scientific institutions. The communicated uncertainty over if, when, and where an eruption would begin led to a reactive plan rather than proactive evacuations. Volcanologists were unable to forecast with certainty how long the eruption would last or if it would restart, again generating considerable uncertainty amongst those affected. However, volcanologists cannot forecast eruptions with any great deal of certainty, as was the case in La Palma, except occasionally at well-monitored volcanoes that frequently erupt in consistent ways, such as the recent eruptions in Iceland. Nonetheless, the 2021 eruption was very well monitored, so in time volcanologists have been able to go back and piece together ‘hindcasts’ of signals indicating the start and end of the volcano’s reawakening, from measurements of lava, seismic signals, ground movement, and gas emissions taken after the eruption had started.
A ‘hindcast’ is an approach where scientists see if they could ‘forecast’ a natural phenomenon e.g. a volcanic eruption, after the phenomenon has happened, once they have had time to fully understand and analyse the event. These help us to identify what signs to look for to understand forecasted future events better.
However, as was the case in La Palma, volcanologists simply cannot forecast eruptions with any great deal of certainty, except very occasionally at well-monitored volcanoes that frequently erupt in consistent ways, such as the recent eruptions in Iceland starting in 2021. Nonetheless, the 2021 eruption on La Palma was sufficiently monitored that, with time, volcanologists have been able to go back and piece together ‘hindcasts’ of signals indicating the start and end of the volcano’s reawakening, from measurements of lava, seismic signals, ground movement, and gas emissions.
Below, we will explore and summarise the academic articles relating to these hindcasts, explaining what’s new, what was hindcast, which techniques scientists used, and how these studies may be useful in the future.
J.Andújar and others, 2025: Lava – replenishment of the magmatic system from 2003-2018.
Scientists have discovered that magma may have been silently moving under La Palma 10-15 years before the 2021 eruption, which could not be detected by their monitoring equipment.
What was measured and why?
Volcanic rocks originate in the mantle as magma and rise through the crust towards the surface where they may eventually erupt as lava. The upwards journey may involve many cycles of heating, cooling, storage in magma reservoirs, and mixing with other magmas. Rather than a traditional singular circular magma chamber, volcanologists now think of volcanic plumbing systems as extensive storage regions across a range of temperatures, depths and chemical compositions.
Lavas are made of many different minerals that are found as crystals in the rock. The crystals of each mineral will have different chemical make-ups, different shapes and sizes, and abundances in a given lava. Each of these crystals preferentially form at different pressures and temperatures beneath the Earth’s surface, and by analysing the abundance, size, and shape of crystals in a lava, volcanologists can reconstruct its journey and infer when and where different processes occurred. This particular study analyses samples from three of the 2021 lava flows to perform high-temperature experiments on them to do this.
What do the results show?
Over the 10-15 years before the eruption, new magma was injected from the mantle into the shallow magma reservoir at around 10 km depth.
Unusually, no earthquakes were associated with these events, which suggests there was already an established connection between the mantle and shallow magma reservoir, and the new magma did not have to force apart any new rocks.
During the eruption, the crustal reservoir emptied by the 27th September, after this date it was recharged by hotter and more fluid magmas from the mantle, explaining the pause and change in style of the eruption after this date.
J. Mezua and others, 2024: Seismicity – evidence of magma migration in September 2021
Scientists used seismic signals to track changes in the subsurface as magma rose prior to the 2021 eruption.
What was measured and why?
As well as typical earthquakes, seismometers can also record continuous very low magnitude signals from background noise, known as ambient noise, from ocean waves and other sources. Seismic waves from ambient noise sources will record information about the layers of the subsurface they travel through, normally in the top few km of the crust. If you measure ambient noise signals across multiple seismometers in an area, with detailed analysis you can generate an image of the layers of the subsurface and how these change over time.
In volcanic eruptions, this has been used to measure changes in the subsurface through time as magma rises to the surface. Compared to solid rock, seismic waves will travel more slowly through hotter and more fluid rising magma, and this can be measured from ambient noise signals. This study applied these methods retrospectively to seismic data from 2018-2022, to see if they could identify any changes leading up to the eruption.
What do the results show?
Starting on the 12th September 2021, a dramatic decrease in seismic velocity was recorded, interpreted as the rapid movement of magma towards the surface.
Prior to 12th September, no changes of a similar magnitude had been recorded.
Whilst this technique was applied retrospectively, there is potential to apply it in future in near-real-time, where it could serve as a valuable volcano monitoring tool.
M. Charco and others, 2024: Ground movement – Evidence for the end of the eruption
Scientists used measurements of ground movement to forecast when the 2021 eruption may end.
What was measured and why?
GNSS stations record the movement of the ground caused by a range of processes. In volcanic areas, the volcanic processes may cause the ground around the volcano to inflate or deflate. These movements are on the order of centimetres, and are often caused by changes in pressure in a magmatic system, causing uplift or subsidence of the ground, respectively. A ‘classical’ but oversimplified cycle for ground movement at volcanoes begins with uplift preceding the eruption until the pressure overcomes the strength of the surrounding rock, causing an eruption. The eruption of material from a magma reservoir reduces the pressure causing subsidence during an eruption, and after the eruption, uplift may begin again if new magma is being pumped into the reservoir.
This study characterises GNSS data showing subsidence throughout the eruption. The subsidence trend was faster at first before slowing to nearly no subsidence by the end of the eruption. The study fit this slowing subsidence trend with a best-fit curve, measuring where the curve levelled off and taking this as an estimate for the end of the eruption. They started with deformation data for the whole eruption, before removing the deformation data from the end of the eruption bit by bit and repeating the study, to see how many days of deformation data they need to get a good estimate of the end date of the eruption.
What do the results show?
This approach was applied during the eruption and provided an accurate forecast of the end of the eruption, but it was too uncertain to be used by authorities during the eruption.
Hindcasts completed after the eruption were accurate and less uncertain, and became more accurate and less uncertain throughout the eruption as more days of GNSS time series data became available.
From around 47 days into the 85-day eruption, the estimates of eruption length from hindcasts became stable and accurate.
For eruptions meeting certain criteria that means this method can be used; forecasts of the end of an eruption may be possible at future eruptions in La Palma and elsewhere.
An example of the type of decreasing trends through time that were seen in the ground deformation and SO2 emissions and were used to hindcast the end of the eruption.
B. Esse and others, 2025: Gas (SO2) emissions – Evidence for the end of the eruption
Scientists use satellite gas measurements to determine when the 2021 eruption was going to end.
What was measured and why?
During a volcanic eruption, gas emissions into the atmosphere, similar to ground movement in the previous example, can be indicative of the pressure in a magmatic system. If an eruption is driven simply by the draining of a magma chamber, such as in La Palma, gas emissions would be expected to reduce throughout the eruption until it stops. A widely used way of measuring volcanic gas emissions is by measuring sulphur dioxide (SO2) in the atmosphere using satellites, as SO2 is easier to detect than other volcanic gases.
This study collected SO2 emission data throughout the 2021 Tajogaite eruption and attempted to use the decay (decrease throughout time) in SO2 emissions to hindcast the end of the eruption. They also attempted to determine what emissions-based threshold could be used to best forecast the end of the eruption.
What do the results show?
They could hindcast the approximate end date of the eruption from SO2 emissions data from as early as the 20th October, with the uncertainty decreasing the as data from more of the eruption as added. For this, they assumed the eruption was over when the gas emissions dropped to 6% of the peak gas emissions at the start of the eruption,
This approach could be used in near-real-time during future eruptions to forecast the end of an eruption, provided the SO2 emissions reduce with time in a similar way to the La Palma eruption.
The threshold of 6% used was the best performing out of the values tested (2-10%), but this may be different for future eruptions.
An SO2 emission map for La Palma derived from satellite data on the 20th October 2021.
My Reflections
The 2021 eruption in La Palma was monitored using innovative techniques and equipment, which have since provided scientists valuable insights into the eruption, including accurately hindcasting when the eruption would end, which would not have been possible to do during the crisis. These advances could improve the ability of monitoring agencies to forecast key events—such as when an eruption might end—in the future. However, in the race to publish findings, many local residents felt that important information had been withheld from them, as they believed this information was “known beforehand”. Moving forward, it is crucial to prioritise effective communication alongside scientific research. Well-monitored eruptions such as this one become hotbeds for new scientific techniques and understanding, but communities may not see these benefits if efforts are not made by scientists to communicate them. By ensuring that local communities understand and benefit from scientific findings, we can strengthen both volcanic resilience and public trust.
No Comments