It’s been a while since my last post in VolcanoCafe. I have been working on multiple ideas and gathering data, but not been able to finish anything. Today, I bring something exciting. I’m using  GIS data (Geographic Information Systems) to explore in depth a volcanic arc, specifically the Aleutian Arc. There’s a lot of potential hidden in GIS data to understand volcanoes, their history, and behaviour. My initial motivation was to get volume estimates of volcanic edifices to better understand how much, where, and how magma is produced, but the project has taken a life of its own and expanded, so we will also be looking at lava fluidity, stress fields, and chemistry. It will be a brief yet profound look into the characteristics of Aleutian volcanism.

The Aleutian Arc

We go to Alaska, amid stormy seas and rugged mountains. Covered in vibrant green grass in summer and white snow in winter, these volcanoes are remote and often hard to access. It is this remoteness that causes the lack of data that many of them experience when compared to other US volcanoes, yet they are still better studied than those in many other places of the world. Ironically, the extreme climate also helps make exposures more available. Past glaciers have cut into old volcanic edifices, revealing piles of stratified lava flows and tephra hundreds of meters thick. Widespread peatland occurrences preserve tephras in carbon-rich material that can help date them in great detail (wiggle matching should be possible, though yet unused technique, in peat sections). These same tephras reach Greenland in trace amounts, even those of subplinian eruptions, where they might one day be dated to the precise year using ice cores. There’s a lot to be learned in the future from it, but in the meantime, a lot can be learned without going to the place.

Pulling and squeezing

Volcanoes often use vertical sheets of magma known as dikes to erupt, which intersect the surface as linear fissures, and these sheets of magma act as compasses of the surrounding stress field, meaning the tension and/or compression the rock is under and their directions. Below is a map showing the direction of all the fissure swarms I’ve been able to spot (shown in red), and also the direction of motion of the Pacific Plate (green), which moves towards the northwest, where it “impacts” the North American Plate and subducts under it.

The fissure swarms, as can be seen, are most of them aligned with the Pacific Plate motion. Why is this? The Pacific Plate is the main player here because it moves at about 7 cm/yr towards the subduction zone, while the North American Plate motion happens at only 3-8 mm/yr, towards the south. It squeezes the arc and forces magma to intrude in the same direction as the pressure is being applied from. If the dike direction (the long axis) were perpendicular to it, then the flat face of the dike (needs space to open) would be oriented towards and directly experience this crushing force, therefore unable to intrude. By aligning their direction to this compressive stress, they can avoid its effects. In other words, the dikes show that much of the arc is being compressed by the plate convergence.

But why do the volcanoes have fissure swarms if there is compression? Many of these swarms are quite weak, but the volcanoes that have more prominent ones are also very prominent in terms of elevation. Shishaldin or Veniaminof, for example, have marked fissure swarms, with abundant volcanic vents, and they also happen to be very tall edifices towering over 2 km above the surrounding area. It’s likely this height that provides the dikes with enough pressure (supported by the magma column under the volcano summit) to rip open the ground.

There is one marked exception in the stress directions, which reaches its maximum expression in the area of Seguam volcano, where dikes rotate to become parallel to the trench. There’s a very different stress field here, one that seems to be caused by none other than rifting. A dike intrusion has to push the ground to the sides, away from it. If a dike’s direction is parallel to the trench, then the ground will have to be pushed towards the trench, yet that’s where pressure normally comes from, so it’s the hardest direction to do so. This is unless the subduction zone is actually experiencing a suction force from the subducting plate at the trench. This suction is created by rollback, the subducting slab dropping downwards into the mantle and retreating away from the overriding plate, but also inevitably dragging the overriding plate with it, pulling it towards it. This creates stretching. Magmatism along the arc often exploits this tension, intrudes into it, and causes the arc to split open, along its length (may eventually turn into a back-arc basin). This seems to be happening at Seguam.

As we advance westwards along the chain, the changes seem to emerge gradually, at the same time as the subduction angle becomes steeper. Westdahl is the first to show a slight rotation in the dike direction. By the time we reach Okmok there are some normal faults in the seafloor, and the field keeps rotating. On reaching Yunaska, a full graben develops along the axis of the arc, which continues as about three graben segments to Seguam, where the field seems purely extensional. This graben complex is known as the Amlia and Amukta Basins, though I think a name like “Seguam Rift” could work better. To the west of Seguam, the setting reverts immediately to compressive.

Locally, other dike orientations exist that are related to the stress fields of inflating magma bodies, most notably Emmons Lake (but also Tana and Vsevidof), which has a fissure system aligned with the long axis of its caldera and magma body. As the magma builds up, it arches the overlying rock, and the shape of the inflating area favors intrusions to run along this axis.

As can be seen in the first map, the movement of the Pacific Plate becomes parallel to the trench to the west of Buldir volcano (last subaerial volcano to the west). With the southward motion of the North American Plate also being quite slow, this area experiences very little subduction and instead is largely a shear boundary. As such, while the full length of the Aleutian Trench is 3800 km, the westernmost 1,000 km have practically no volcanism. Some minuscule volcanic fields can be seen in the seafloor between Piip volcano and Buldir, often referred to as western cones.

Piip volcano stands lonely and mysterious near the west end of the Aleutian Trench and at the south end of the so-called Komandor Basin. This whole basin is formed by very recent seafloor spreading that was active until only 10 million years ago, according to current ocean crust age models. However, Piip at the southern end of this system is an active stratovolcano (has fumarolic activity and recent volcanic deposits), and also features some young-looking spreading ridge structures under and around it. Too young-looking to be anywhere as old as 10 Ma. I believe the Komandor spreading ridge still continues to be active under Piip. A junction between a feeble volcanic arc and a dying ridge. It probably accommodates some rollback exerted by the Aleutian Trench. The rest of the former Komandor spreading may perhaps have come from the now-dead subduction rollback along the nearby Shirshov Ridge.

Piip is an unusual volcano, not just due to its isolated location and arc-spreading combination, but also due to its strong high-Mg andesite chemistry that I will mention later, and this draws some parallels to the West Mata volcano of the Tonga Arc that is in a similar setting and erupts similar lava.

Eruption rates and viscosity

I’ve analysed the eruption rates and viscosities of volcanoes from Buldir eastwards. They comprise a total of 50 volcanic systems, spaced at mean intervals of 52 km. That said, I’ve treated volcanic complexes as one system that GVP often lists as multiple volcanoes. These two, eruption rate and viscosity, are fundamental parameters of volcanism related to the processes that give rise to the volcanoes and also fundamental factors controlling the volcano’s behaviour. And yet they are rarely ever given the importance that they have, probably because it has previously proven difficult to obtain estimates. Even now, with the wealth of GIS data that exists, what I’ve done is partly an art rather than a precise science, but I think nonetheless I’ve been able to apply a consistent technique that is quite useful and close enough to the actual values.

Isn’t chemistry a good indicator of viscosity already, one for which there is plenty of existing data? Not really; while viscosity is indeed mainly controlled by the silica concentration in the melt, most existing chemistry data refer to the lava as a whole, which is not the same thing. Magmas/lavas include liquid, gas, and solid. They contain bubbles of volcanic gas that mostly escape during an eruption, but there are also crystals in the magma, often 50% if not more, that stay with it even after erupting. These crystals will have a lot less silica than the melt and will substantially shift the overall composition towards mafic. Andesites and basaltic andesites are the dominant lavas in the Aleutians, but a lot of these mafic-intermediate volcanoes have a melt fraction that is actually rhyolite, so they form viscous lava domes and coulees. That’s why I’ve looked directly at the products of the volcanoes to obtain their true melt viscosity and approximate silica content.

What I’ve done is measure the minimum lava channel width of lava flows; this is the minimum width for lava to form a free-flowing stream and is a rough indicator of viscosity. Fluid lavas may form free-flowing streams that are only tens of centimeters wide in places like Hawaii. On the other extreme, at their most viscous, lava channels will be no narrower than 200 meters. Using cases with reported glass silica content, I was able to get a rough relation between the two parameters, which is the following:

Melt silica%=48.438*Minimum lava channel width^0.0903

The SiO2 can be off by up to 8 wt%, but it’s still quite useful to get approximations of the silica content of the melt fraction of the magmas. There is some offset because the way I’ve estimated viscosity already is a rough one, and also because there will be some factors other than silica that may affect the viscosity of the lavas. 200 m channel width is sort of a limit because at this size the above equation yields 78 wt% silica, which is the most silica magma is known to reach. If I get channel estimates of over 200 m, or find volcanoes too viscous to flow at all (domes surrounded by PDC aprons), I assume an equivalence to 200 m and thus 78 wt% silica.

I obtained eruption rates by estimating the volumes of edifices younger than about 1 Ma, using their state of erosion, and helped by existing age data of volcanic products. Taking an interval this long helps ensure that two or so shifts in the locus of volcanic activity have taken place since, and thus separates edifices by their ages and also helps approach a geologic long-term, if such a thing exists. I filled glacier valleys to account for their powerful and relentless erosion.

There were some problems, however, when it came to pyroclastic shields. When a caldera covered in glaciers collapses, the ignimbrite produced immediately starts melting the ice, which creates enormous lahars that carry the pyroclastic material all the way to the sea. For this reason, it seems the frequently glaciated recurrent calderas have been unable to build substantial edifices. Powerful volcanoes like Fisher, Okmok, and Semisopochnoi have actually lost elevation over time, as glaciers dissected their long-lived shields and explosive caldera-forming eruptions were unable to deposit ignimbrites on the thick ice cover. I had to make guesses for the volumes of these three volcanoes based on nearby systems that seemed similarly vigorous. For Emmons Lake, I used the five known ignimbrite deposits and approximated their volumes based on the caldera size (100-150 km3 dense rock equivalent each, using the DRE for the volume estimate to try and keep consistency, because DRE often falls close to the ignimbrite/PDC volume, and it is this last component what I’m measuring in general with other volcanoes since distal ash won’t contribute to edifice growth).

The graphs below sum up the results of the investigation.

Overall, there’s probably some underestimation since I’m not counting the distal ash (which doesn’t contribute to volcanic edifice growth), but I don’t think the number would change too much when including it. In the last 1 million years, the Aleutian Arc seems to have erupted 12,000 km3 of volcanic material. This is equivalent to about 12 km3 every 1,000 years, averaged over the considered period. Is this a lot? I only have a similar estimate for the Lesser Antilles Arc: 1.3 km3/1000 years. True, the Antilles are a lot shorter, but still, when dividing by the length of the arc, the Aleutian Arc comes out to be more than twice as productive per unit of distance. Though other arcs like Sunda or Kamchatka are definitely more productive than Aleutian. The Aleutian Arc eruption rate would be about half that of the Reunion hotspot, and only a tenth or less that of Hawaii, but likely a lot stronger than many weak hotspots.

When taking viscosity and eruption rates together, it becomes clear that the Aleutian Arc has some substantial variation in the way it performs volcanic activity. And this variation almost seems symmetrical around a point near the center of the arc. There’s a central segment where the eruption rate is generally higher, and lavas are fluid. This central segment more or less goes from Shishaldin to Atka. A lot of volcanoes in this area are recurrent caldera systems, and as we will see, they also have a characteristic chemistry. To the sides of the central segment, we can distinguish eastern and western segments that run all the way to the extremities of the arc. The main difference is that these two lateral segments erupt more viscous lavas, often lava domes or thick coulees, and are more drawn towards building stratovolcanoes with a single central vent. The volcanoes in the flanking segments are still dominated by basaltic andesite and andesite lavas like the ones in the central part, but they are more often crystal-rich; therefore have a more evolved and silica-rich melt fraction. The other difference is that lava eruption rates tend to drop from the center towards the extremities of the arc. This is particularly true for the west end, where volcanism is already quite feeble on reaching Buldir, but a wisp of volcanic activity still continues to its west, fading into weak submarine fields.

In the eastern segment, which spans the Alaska Peninsula, there’s a curious behaviour. Five volcanic systems concentrate almost the entire lava output: Hayes-Spurr, the Katmai Group, Aniakchak, Veniaminof, and Emmons Lake. These five systems vary in their eruption style, but they do have in common that they seem to have a more mature storage (either shallow chambers or more expansive areas of magma storage in the crust). True, this could be a consequence of the higher supply, yet it could also be the cause. The big five may, thanks to their greater development, have a stronger pull on magma production and starve the weaker systems around them. This volume disparity seems to strengthen eastward across the entire arc, likely related to the growing continental crust thickness. Emmons Lake yielded the largest lava production among the volcanoes in the arc, 1250 km3 in the last 1 Ma.

The Ancient Aleutians

Having calculated the 1 million years lava output creates a nice opportunity to compare it to the magma production over the entire history of the arc. The arc is young; the continental crust grown on top of Cretaceous oceanic crust, from Shishaldin westward, belongs to magma production along the current subduction zone. Because the oldest ages of volcanic arc products are about 47 Ma, it’s usually assumed that’s the moment when the arc first initiated. I used the Gareloi-Shishaldin section for the comparison. Turns out the total magma production since inception has been 3 million km3. This is a lot more than I was expecting to get, since it would imply rates of 71,000 km3/Ma over the last 47 Ma, compared to the 6,500 km3 this section has actually seen in the last 1 Ma. There’s definitely an important intrusive volume below the surface contributing to the continental crust growth, but it shouldn’t do much more than double the number, so that, even then, the gap would be too great.

The answer to this problem, I think, is that the arc may have actually started earlier. As it grew, the older products were probably buried by younger volcanism and subsided as material piled up on top. I think it works better, and also makes more sense tectonically, if the Aleutian Arc started with the opening of the same back-arc basin that created the Bering Sea oceanic seafloor, so about 85 million years ago, during rollback of the former Kula Plate. Would be a similar mechanism to how younger island arcs like Tonga, Vanuatu or the Marianas have formed; older arcs that split into seafloor spreading. Subduction rates must have been higher back in the day, and the Cretaceous was quite skilled at producing powerful volcanism, so this Late Cretaceous Arc may have been more productive than the current one and may explain the huge amount of continental crust.

Chemistry

Lastly, we will be going over a chemical review of the Aleutian Arc. There are two aspects I’ve found interesting to look at: silica concentration per given MgO and TiO2-Fe2O vs Al2O3-CaO per given MgO. The MgO is always a good way to judge how primitive a magma is. This is because, no matter what crystals are coming out of the magma, the MgO is always more concentrated in these crystals and depleted in the melt. Magma will always travel to less MgO as it cools and evolves. Also, evolution tends to end around the time MgO hits 0 and start around 10 wt% or more for alkaline magmas.

Both aspects I’ve looked at seem related to each other, as well as to the viscosity and eruption rates, but the exact reasons as to why this is seem unclear. In general, I think that a lot of aspects of volcanoes remain poorly understood.

SiO2 vs MgO shows that no volcanoes are silica-undersaturated relative to the subalkaline volcanoes of places like Iceland, so although some volcanoes will fall in the alkaline field of the TAS diagram, I think when looking at the silica content, none of them really have the low silica characteristic of alkaline magmas. Silica per given MgO increases from a low at the center to a peak at both ends of the arc. The volcanoes near the sides erupt strong high-Mg andesite/calc-alkaline magma. On one side, we find this magma at Buldir and the submarine fields to its west, plus Piip volcano, and on the east, we find it at Douglas-Fourpeaked, Augustine, and Hayes-Spurr. High-Mg andesites are magmas that have an inherently high silica content. In the more extreme cases, they may already form in the mantle with silica contents of 57 wt% SiO2 (between andesite and basaltic-andesite) while featuring high MgO comparable to primitive basalts.

Something very interesting, too, is what I call anorthite-pyroxene fractionation. I name it because I don’t think it has a name, and, if it does, I’m not aware of it. Basalts (and low-SiO2 basaltic andesites) that are stored at low pressures can undergo a specific kind of crystal fractionation in which the magma evolves while barely increasing in silica. Sometimes, over half the magma can crystallize, while the remaining half will only have risen by ~2 wt% SiO2 and remain almost as fluid. This kind of crystallization is probably what allows lava lakes that pond inside craters in Hawaii to break out downslope decades later and still erupt very fluid lava (like the Keaiwa 1823 flow). During the time this happens, the crystallizing minerals are about half of them pyroxene and half plagioclase.

Anorthite-pyroxene fractionation leaves a marked chemical imprint in that, as the magma evolves, Fe2O3 and TiO2 become enriched (unlike in other forms of fractionation), while CaO and Al2O3 are depleted. The main culprit of this process is probably the calcic variety of the mineral plagioclase, known as anorthite, with the formula CaAl₂Si₂O₈. Anorthite is more stable at lower pressures; it’s clearly what depletes the calcium and aluminium, and probably also helps increase the iron and titanium by displacing other mineral types that would take them.

I’ve used the chemical data available to sort most of the volcanoes into different “iron” series depending on the amount of pyroxene-anorthite fractionation they’ve seen. Low Fe volcanoes will have seen no or very little. The following is the list of volcanoes and the group that they belong to; the numbers in parentheses are the melt silica contents, obtained earlier from the minimum lava flow viscosity.

Low Fe series: Spurr (66), Augustine (75), Douglas-Fourpeaked, Buldir (78), Ingenstrem Depression, Western Cones

Medium Fe series: Redoubt, Kaguyak, Snowy Mountain, Katmai Group (65), Yantarni (78), Emmons Lake (63), Amak (74), Frosty (78), Bogoslof, Rechesnoi-Vsevidof (62), Atka (57), Koniuji (77), Great Sitkin (76), Adak (78), Kanaga (68), Bobrof (78), Tanaga (64), Little Sitkin (75), Kiska (73).

High Fe series: Aniakchak (61), Veniaminof, Shishaldin (53), Fisher, Westdahl (60), Akutan (61), Makushin, Okmok (58), Seguam (55), Semisopochnoi (63).

I found the results to be surprising. Once again, the differences between the central segment and the side segments emerge. The volcanoes with a strong high-Mg andesite chemistry also have low Fe; this may be because their inherently high silica or other chemical characteristics block anorthite-pyroxene fractionation altogether.

But what’s most interesting is that the high-Fe series volcanoes, all of them except Shishaldin and Westdahl, have collapsed calderas during the last ~13,000 years. What’s more, some of them have done so twice. Aniakchak, Veniaminof, and Okmok have each collapsed twice during this time, and their calderas are all about 9 km across! And even better, the areal-largest nine Aleutian caldera-forming events (those at or beyond 4 km diameter) of the last 13,000 years have ALL come from the few high-Fe group volcanoes. Why is this?

Well, while I don’t know all the details of the anorthite-pyroxene fractionation, it seems to benefit from low pressure (shallow magma storage) and basaltic melts. I speculate that the volcanoes that belong to the high Fe series have basaltic/basaltic-andesite melts reaching their shallow chambers, only a few km below the surface, where they can crystallize in this style for a long time and undergo the corresponding chemical changes. The primitive Al-rich basalts rarely make it to the surface, except at Seguam. This low-silica melt reaching their shallow chambers will be highly fluid, similar to volcanoes with known lava lake activity, like Masaya or Ambrym, and will thus allow magma convection even in relatively narrow conduits. As such, convecting mafic melts between the asthenosphere and the shallow storage of these volcanoes, I think, could be helping sustain magma chambers that stay molten thanks to this steady supply of heat. And thanks to their thin roofs, collapse easily, producing caldera-forming ignimbrite eruptions at short intervals.

The ignimbrites are sometimes mafic-intermediate (Veniaminof, Okmok, Akutan, Makushin and the western caldera of Seguam), sometimes silicic (eastern caldera of Seguam and Semisopochnoi), or sometimes both (Aniakchak and Fisher). As another reflection of their mature caldera style, some of these volcanoes have a ring of vents within their caldera along what must be the location of their outward-dipping ring faults. A heavily intruded ring fault. Their edifices can be lava shields, with massive piles of lava flows coming from flank and summit vents (Veniaminof and Westdahl), or almost pure pyroclastic shields with an output that largely goes into the caldera-forming eruptions (Aniakchak, Fisher, Okmok, or Semisopochnoi). The central segment, understudied and data-lacking volcanoes of Amukta and Yunaska, I suspect, also to be in the high-Fe group if samples were available, and Yunaska is definitely a Holocene caldera-forming eruption producer.

Concluding remarks

The Aleutian Arc is a mid-productivity arc. Volcanism seems organized into an interesting structure that changes from the center to the sides. The central area has higher eruption rates and more fluid lava. I speculate that many of the central systems have long-lived shallow magma chambers and convection of fluid magmas into their base, which allows them to become recurrent sources of caldera eruptions. Eruption rate decreases towards the edges and gradually fades into nothing towards its western extremity.

Volcanoes are varied. I haven’t had a chance to look much into each of them separately, but I may do a future article doing so. There’s still much to learn from them. I hope you all enjoyed this short but dense overview of its volcanism.