There is no stronger evidence for melting of portions of the crust and mantle than the fact that volcanoes exist on the Earth. Volcanoes involve ascent of molten rock (magma) from deep in the crust and upper mantle, and the process of melting is key for ongoing chemical differentiation of the Earth. The materials that ultimately reach the surface, and erupt from a volcano are strongly modified from the original material that melted, with only the lighter components, enriched in volatiles, reaching the surface. The rocks that form from volcanoes are thus different than the original rocks that melted to give the magma, and some of these rocks are so buoyant that they have progressively added to the Earth's continents, while oceanic floor is recycled back into the mantle.

So, melting occurs inside the crust and mantle. Yet, seismic S waves propagate through the crust and mantle, so it is not everywhere molten. In fact, melting occurs only in localized places, and requires somewhat special circumstances. Any rock material will melt, meaning to transform from solid phase to liquid phase, when the temperature exceeds the melting temperature of the rock. The melting temperature varies with pressure (generally it increases), so the temperature tends to only exceed the melting temperature over a localized depth range.

So, the temperature structure of the Earth is key to understanding the occurrence of melting. First we must ask why is the Earth hot inside to begin with? This takes us back to the discussions in the first third of the class, where we talked about the huge amount of heat resulting from accretion of the Earth, as planetisimals collided and built up the mass of the planet. That heat is called primordial heat, as it stems from the Earth formation itself. Associated with the earliest phase of the Earth, there was additional heat released when the iron sank to make the core. This release of gravitational energy gave additional heat to the early Earth, which is slowly dissipating, but has not yet escaped from the planet. The other source of heat for the Earth has been a continuously generating heat source, involving the decay of radioactive elements. Recall that the Earth formed with a distribution of very heavy, unstable elements, which have subsequently been undergoing spontaneous fission processes to decay to more stable forms. While the radioactive fuel is slowly being burned up, and there was more heat production early in the history of the Earth, there are still large amounts of radioactive materials continuing to heat the interior.

OK, given these three main sources of heat, how hot is it inside? This is actually not all that well known, as there are not many observations that we can make at the surface that tightly constrain the temperature at depth. Generally, we do know that the temperature increases with depth almost linearly through the lithosphere, down to depths of 75-100 km under oceans and 200-300 km under continents. In the lithosphere, which is stiff and does not flow, heat is transported by conduction, the transfer of heat by atomic vibrations without mass transport. At the base of the lithosphere the temperature has increased to on the order of 1000-1200 degrees, and if it continued to increase linearly, it would intersect the melting curve of mantle rocks which is about this temperature. But, as the temperature increases the rheology of the rocks changes, and they become lower viscosity and more deformable. At temperatures of around 1000 degrees, the rock becomes soft enough that it is easier for it to move plastically to transport heat out than it is for conduction to transport heat out. The efficient transfer of heat by flow, or convection, causes the rate of temperature increase with depth to greatly diminish, and through the rest of the mantle the temperature gradient is rather small, increasing the temperatures to about 2500 degrees near the base of the mantle (2600 km deeper than the lithosphere). It appears that the mantle rock melting curve is at higher temperatures throughout this region, so there is little or no melting.

At the base of the mantle the temperature again increases rapidly with depth, in the thermal boundary layer between the core and the mantle. This boundary layer again involves heat transport by conduction, as there is no mass flow across the core-mantle boundary. The temperature right at the base of the mantle is somewhere from 3000-4000 degrees. Within the outer core, which is vigorously convecting, the thermal gradient (geotherm) is small, with maximum temperatures of 5500-7000 degrees at the center of the Earth. One of the key tie points that we try to determine experimentally is the melting temperature of iron at pressures corresponding to the inner core-outer core boundary, as this transition from liquid to solid corresponds to the geotherm intercepting the melting temperature of iron (the geotherm is above the melting temperature in the outer core, and below it in the inner core. Unfortunately, the problem is complicated by both the difficulty of replicating the enormous pressures at depth and by uncertainty in the light alloying components of the core, which can strongly affect the melting temperature of the alloy.

So, there are perhaps 3 regions where melting occurs in the interior. One is in the outer core, which is fully molten, or at most a slurry with few particulates. Another region is possibly in the rapid temperature increase at the base of the mantle. There is now seismological evidence suggesting partial melting of the mantle in the core-mantle boundary vicinity in localized regions. The most important melting occurs up near the base of the lithosphere, where the geotherm increases temperature rapidly to the vicinity of the melting curve.

To understand the shallow melting processes better, we will consider the three basic environments in which melting takes place. The tectonic regions with volcanoes include:

1. Subduction Zones (Island arcs and Continental arcs)

2. Rifts, Mid-ocean Ridges (Extensional regimes)

3. Hotspots (Volcanic centers removed from plate boundaries)

The best known volcanoes are in Subduction Zones, in part because these tend to be explosive, deadly volcanoes. But some rift volcanoes are well known (Mt. Kilamanjaro), as are hotspots (Iceland, Hawaii, Tahiti). If we consider the major volcanic eruptions of the world, 15% are in continental rifts, 80% in subduction zones, and 5% in hotspots, but these numbers apply only to the subaerial eruptions. Far more volcanic activity occurs under the oceans.. There are from 15-20 large subaerial eruptions each year.

The subduction zone volcanoes include many of the famous names: Fuji, Mt. St. Helens, Vesuvius, Pelee, Katmai, Krakatao. These volcanoes are distinctive in being the classic, conical shaped mountains built up of layer upon layer of lava flows, in what are called strato-volcanoes. Mt. Fuji, which last erupted in 1707 is a classic volcanic cone, one which you can hike up in a few hours (usually at night so that you can catch the sunrise, if you are lucky and have a clear morning). Yet it is an active volcano which we fully expect to continue to erupt again and again in its long geological history. Mt. Fuji is located over the subduction zone where the Pacific plate underthrusts beneath Japan, and it owes its existence to the melting process that occurs in this convergent zone. The Aleutians, Tonga, Marianas, Kuriles, and other island arcs have similar volcanoes. Subduction zones along continental margins also result in dramatic strato-volcanoes, such as in the Cascades and the Andes. But, why are there volcanoes in these regions?

This question has puzzled geologists for a long time. The subducting oceanic lithosphere in a subduction zone is as much as 1000 degrees colder than the surrounding mantle near depths of 100-200 km. Thus, it is well below the melting temperature for basaltic rocks that make up the oceanic crust, and this is true for the depleted peridotites that make up the rest of the oceanic lithosphere as well. Clearly, volcanoes in island and continental arcs are not caused by melting of the oceanic plate as it sinks into the mantle. The surrounding mantle is actually cooled by the presence of the sinking plate as well, since material is dragged down along with the plate, thus displacing colder material from shallow to deeper depths. So, what is the cause of the melting that feeds volcanoes such as Mt. Fuji?

The answer lies in understanding the sensitivity of melting temperature of rock to the presence of volatiles such as water. If a small amount (say 2-5%) of water is added to a rock, the melting temperature at a given depth can be reduced by hundreds of degrees. Thus, rather than explain the melting as a result of locally higher than average temperatures, we account for the melting by locally lowering the melting temperature by adding water to the rocks. Water is bound up in hydrous minerals in the sediments that accumulate on oceanic plates, and there is actually water that circulates down into the oceanic crust. When the plate sinks into the mantle, the water is transported downward. As the plate heats up slowly the hydrous minerals can become unstable, releasing the water in dehydration reactions. This water is then expelled from the slab and leaks out into the overriding wedge of material, lowering its melting point. At depths of 100-150 km, where the water is expelled from the slab, melting occurs due to the lowered melting temperature (despite the cooling effect from the downwelling slab). The melting reduces the density of the molten rock and it then rapidly ascends, feeding into magma chambers and conduits that build up pressure and eventually erupt in arc volcanoes.

As the magma melts, ascends, ponds in magma chambers, and eventually erupts, it can chemically differentiate significantly, so that the rocks that form from the magma bear little resemblance to the mantle in the wedge that melted. Some of the changes include interactions with the overriding crust through which the magma must traverse so that it can erupt. The most common rock type in island arc volcanoes is andesite. This is a rock that is very sticky, the result of a high silica content. The stickiness, or high viscosity, allows the layers of flows to build up a large cones. It also causes the volcano to plug itself up, which allows pressure to build up between eruptions.

As a result, arc volcanism tends to be explosive. This is the combined result of high silica content rocks, which are sticky, and the presence of lots of volatiles and gases in the magma. Arc volcanoes tend to have long periods of repose, with eruption intervals of 100-1000 years. This time scale makes it challenging to anticipate future violent eruptions, as we typically have very limited history of eruptions to extrapolate from. The source magmas of arc volcanoes are rich in water and gases both due to volatiles that were carried down by the subducting slab, and due to interactions with the overriding plate as the magma ascends.

Arc eruptions are an important part of the process of continental growth, as they extrude large masses of chemically differentiated rocks. Mt. St. Helens erupted about 0.5 cubic km of new rocks, but that is a modest amount for a large volcano. The 1883 eruption of Krakatao ejected 6.0 cubic km, while the 1912 eruption of Katmai gave out 12.0 cubic km. But far greater eruptions have happened in the past. The explosion that produced the Long Valley Caldera gave out about 600 cubic km. Ash from this event spread thick over states as far away as Nebraska. A repeat event would be devastating for any country, and perhaps for the entire hemisphere in which it took place. To put the volcanic energy release into perspective, the violent 1980 eruption of Mt. St. Helens gave off energy equivalent to 27,000 nuclear bombs like that dropped on Hiroshima. This would involve 1 such explosion each second for 9 hours.

A caldera is the large ring-like collapse structure that results when the surface layers overlying a massive magma chamber collapse in, as the chamber empties during a big eruptions.

While arc volcanoes tend to get most of the public recognition due to their explosive behavior and the intermittent human catastrophes that result, they are a relatively minor component of the Earth's overall volcanic budget. Rifts, or spreading regions, are zones of massive volcanism, much of which occurs with little fanfare because the eruptions are under the oceans. Indeed, the entire, global-encircling chain of mountains along the mid-ocean ridge system is a vast chain of volcanoes, built up by the up-welling molten material that fills in the gap in the spreading oceanic crust and lithosphere. The present day rift system extends the entire length of the Atlantic, through the Indian Sea, where it bifurcates into a chain extending up into the Red Sea and then southward into the East African rift system (a continental rift that connects up to spreading oceanic ridges) and a separate chain that connects over into the East Pacific rise system.

Thousands of kilometers long, the mid-ocean ridge volcanic system accounts for 90% of annual volcanic activity. Of all subaerial volcanoes, only 15% are rift volcanoes, such as Mount Kilamanjaro, in the East African rift, or the volcanoes of Iceland, but these do contribute much to continental volcanism. However, the great volume of rift volcanism is in the Mid-Ocean ridge system, with 2.5 square kilometers of new seafloor area being produced each year, and a corresponding 12 cubic kilometers of magma injected into the spreading rifts. The average spreading rates of most ocean rifts is about 15-20 cm/yr, so you can imagine a fissure that wide opening each year along the zipper-like extent of the entire mid-ocean ridge system, filling with hot magma that cools to form new oceanic crust. This process has produced over 60% of the Earth's surface, and completely repaves this region every 200 million years or so.

On the basis of a few direct observations of submarine eruptions, and many more detections of eruptions from the underwater hydroacoustic system operated by the U.S. Navy to monitor submarine and ship activity, there are an estimated 20 eruptions each year on the mid-ocean ridge system. These eruptions tend not to be explosive, and involve "effusive" eruptions of runny magma. In part this is the result of eruption under a thick layer of water, some 3 km deep, the pressure of which keeps gasses and fluids in solution with the molten rock rather than letting the gas separate out and build-up pressure.

With so many submarine eruptions, it is no surprise that much of ocean water chemistry is the result of interactions of erupted products with the overlying water. Gases and fluids brought to the surface have in fact produced the atmosphere and the oceans, and continue to contribute to the fluid envelopes of the Earth.

A cross-section through a mid-ocean ridge system would show a complex layered structure in the upper few kilometers under the water layer. There is commonly a shallow magma chamber, perhaps a few kilometers in diameter down 2-5 kilometers under the center of the ridge. Above this chamber there is a sequence of vertical tabular dikes, or intrusions of molten rock along vertically oriented cracks. The sheeted dike sequence is overlain by pillow lavas, a layer of bumpy extruded rocks that cool quickly in contact with the ocean water and fill up from below with magma prior to cooling or bursting to make pillow upon pillow. Below the magma chamber there are horizontal layers of rock which result from settling out of refractory crystals in the magma chamber.

So, why is there magma in this region? What has caused the localized conditions enabling rock to melt? The general explanation is that due to the rifting process, the hot interior of the Earth can ascend more rapidly than would normally be the case. As the hot material rises, it encounters decreasing pressures. Material can melt if it rises fast enough by a process called decompression melting. The basic idea is that as pressure increases on a substance it has to be heated to larger and larger temperatures in order to melt. Thus, the melting curve of rocks increases with depth in the Earth, even for uniform rock composition. We also know that the temperature increases with depth, although for most of the mantle the temperature does not increase to higher than the melting temperature. But if we take very hot rock, which under high pressure would still be solid (sub-solidus), and we abruptly lower the pressure (bring it to shallow depth) without letting the rock cool off, the decrease in melting temperature due to decreasing pressure may allow the rock to melt. Thus, rapidly upwelling solid mantle (moving by solid state deformation or convection) can intersect the melting curve and partially melt under the ridge. This melt separates from the solid matrix and accumulates in magma chambers or large cavities filled with molten rock. Because the melt chemically differentiates (remember, every rock involves multiple minerals, each of which has a different melting temperature: heavy, refractory crystals don't melt and separate out as solid dregs, while the material that is above its melting curve tends to rise because it is less dense than the residue) the molten component that rises to the surface is quite different in overall composition than the original rock that melted. This melt fills the gap as the plates spread, and the cooling rock is incorporated into the growing plates on either side of the rift.

In most cases, it is believed that sea-floor spreading is actually a rather passive process, meaning that the plates are not forced apart by the upwellings, but rather the upwelling occurs because the plates are pulled apart primarily by the old subducting slabs at the other edge of the plate. In some cases, though, upwellings can break apart continents, as is happening in eastern Africa, and as broke South America off from Africa. The upwellings then evolve into spreading ridge systems like the mid-Atlantic rise. The ridges themselves are defined as ridges because they are hot regions which tend to be buoyant. As the ocean crust and lithosphere increase in age (in the perpendicular direction to the ridge) the rock cools, contracts (densifies) and rides lower on the surface of the Earth. There is thus a systematic increase in ocean depth with distance from the mid-ocean ridges, caused by cooling of the oceanic lithospheric plate. This persists up to the point where the lithosphere is cold enough to become gravitationally unstable, whereupon it sinks into the mantle on its own, pulling the trailing plate of younger age downward behind it.

So, how do we know about the deep structure of mid-ocean ridges? There are recent efforts to go explore these ridges in situ, using ships that conduct seismic experiments. Using underwater explosions or air bursts, P waves are sent out from source and bounced off of the subsurface layers, ultimately being recorded on seismometers towed in very long cables behind ships. Another source of information is the chemistry of the rocks themselves, which reveal the nature of the melting, differentiation, and ascent processes. More direct evidence is often provided by rare instances of exposed oceanic crust that was upthrust onto continents (often during continental collisions) rather than being subducted as is the fate of most oceanic crust and lithosphere. There are a few places where old oceanic crust is found exposed on the continents, in formations that are called ophiolites. This includes sites in Cyprus, the Western U.S., Oman, Greece, Papua New Guinea and Noumea. These regions show the pillow lava layer, the sheeted dike complex, and the deeper cumulate layers produced by refractory materials that have separated in the magma chamber. Very consistent structures are found in different regions suggesting that the process of oceanic crustal formation is fairly ubiquitous for all mid-ocean ridge systems.

If we go out to any oceanic crust and pick up a rock, it will be made of basalt. This is a type of rock that is about 50% silica, with relatively enriched magnesium, iron and calcium compared to most continental rocks. Typically a fresh (unweathered) basalt is pretty dark or black, due to the presence of the heavy metals. Basalts are the typical rock type that will result when any chunk of the mantle is melted and the lighter material separated and then cooled. Since most of the Earth's surface is covered by basaltic ocean floor, it is the most common rock found on the surface of the Earth.

In contrast to andesitic rocks that are erupted in arc volcanoes, molten basalt is a low viscosity, runny magma type. This is because basalt is relatively low in silica (which makes andesite magmas sticky). As a result, basalt builds up very broad volcanoes called shield volcanoes, with layer upon layer of flows and underground injections. The viscosity is about the same as that of chunky soups, so steep-sided volcanic edifices cannot be built up by basaltic eruptions. This runny nature of basalt magmas is also a factor in preventing rift volcanoes from building up large pressures for explosive eruptions. The volcanoes tend not to plug themselves up with sticky rock nearly as effectively as in arc volcanoes. Combined with eruption under water, which causes the gasses and other volatiles to remain in solution in the rock, effusive eruptions result.

As new oceanic crust is formed by cooling of pillow basalts and deeper dikes, it is strongly fractured by joints and faults, some of which are caused by earthquake deformation due to the crustal extension, and some of which are the result of thermal contractions associated with cooling. Water in the ocean can seep down into these cracks and as it penetrates into the new oceanic crust it is heated up. The hot water then interacts with the rocks, leaching minerals and elements out of the rock. The hot water increases in temperature and then rises, carrying this bounty of leached elements with it, and jets out onto the ocean floor in vents. Some of these vents, first discovered by submarine dives about 20 years ago, are so enriched in heavy materials such as copper, zinc, lead and sulfides that the hot steam and water produces 'black smokers' or vents of darkened hot water and steam 'erupting' at the base of the ocean. Much of the material leached from the rocks settles out onto the ocean floor (producing economic concentrations of some materials such as manganese nodules), but some is mixed into the ocean, controlling the ocean chemistry.

In addition to the vast upwelling regions along ocean ridge systems, there are other major volcanic centers that are removed from plate boundaries, or only incidentally associated with them. This includes major mid-plate ocean island volcanoes such as in Hawaii, Tahiti, or Fiji. In the Pacific Ocean in particular, there are long chains of islands that trend NW-SE, with older, more eroded islands toward the northwest. It is believed that many of the island chains have been produced by the northwesterly motion of the Pacific plate over several relatively fixed locations of melting below the plate. These 'hotspots' appear to remain fixed relative to the plate because their origin is deeper than the plate itself, with up-wellings that may come from as deep as the lowermost mantle. This causes us to view hotspot volcanoes as somewhat distinct from the rift and arc volcanoes which are clearly related to the creation and destruction of oceanic lithosphere.

Effectively, the up-welling hot rock in a hotspot burns a hole through the overriding lithosphere, whether it is oceanic or continental, and the magma produced by the hot temperatures of the up-welling (originating by the same type of decompression melting as occurs under rifts) penetrates to the surface, producing a volcanic mountain superimposed on the plate. The motion of the plate translates the mountain in the plate motion direction, eventually decapitating the magma conduits that fed into the mountain, and shutting off the volcanic activity. At that time, the magma burns a new path up to the surface, creating a new volcano in the chain. This is much like passing a sheet of metal over a fixed blowtorch, creating a welded scar in the sheet due to passage by the heat source.

The most famous hot spot island chain is the Hawaiian islands, which extend all the way from the current hotspot site under the big island of Hawaii, out to the northwest past now extinct volcanic mountains of Maui, Oahu, Kauai, Midway, on to submarine atolls along the chain and on to submarine mountains along the Emperor Island chain, eventually leading to Kamchatka where the Pacific plate subducts under the Eurasian (in detail under the North American) plate. The age of the islands increases toward the northwest, and erosion of the exposed island and submarine landslides of the deeper underwater mountain whittle away the islands until they are below the water. While the three volcanoes on Hawaii (Mauna Loa, Mauna Kea, Kilauea) are all active volcanoes (Mauna Kea is dormant, possibly permanently, but probably not; many large telescopes have been built on its summit with the hope that it not erupt again soon....). A new island is building up on the southeastern flank of Hawaii (Loihi), and this may grow into the next island in the chain. Because the ocean depth is 4 km, the exposed mountains on the islands are really the crest of giant volcanoes, with Hawaii being the largest mountain on the Earth, if we define the structure relative to the surrounding sea floor level. This reflects huge outpourings of lava, but recognizing that Hawaii is just the youngest of many volcanoes produced by the same hotspot, we can begin to appreciate the magnitude of melting associated with this hotspot.

The large, sustained melting under an island chain requires a steady heat source persisting for many tens of millions of years (the oldest islands in the Hawaiian-Emperor chain are almost 80 million years old, and yet older ones appear to have been subducted away from the surface). Many Earth scientists believe that plumes of hot up-welling material are responsible, with a plume that may be 50-100 km cylinder of rising material extending possibly as deep as the core-mantle boundary. The rock that melts under hotspots has some distinctions from mid-ocean ridge basalts, but this is only true of the minor components. The main rock types of hotspots found in the oceans is again basalt (the typical produce of melting mantle rocks), so the whole island of Hawaii is a massive pile of basaltic lava flows. But minor chemical differences reflect the deep-seated origins of the upwelling material, which differs from mid-ocean ridge materials. The basalt flows on Hawaii have two distinct types. The Hawaiian names are Pahoehoe and Aa. Pahoehoe is like road tar, with fluid, but often ropy or rippled surfaces. Aa is more broken up and blocky, and the lava flow tens to be meters thick.

Since we think that hotspots are generally independent of the shallower mantle circulation directly involved in the production and destruction of oceanic lithosphere, we would expect up-welling plumes to produce some hotspots below continents as well. Indeed, there are similar 'hotspot' tracks burned into the continental rocks of all of the continents. In North America, the best example is the Yellowstone hotspot. The westerly motion of the North American plate has pushed the plate over a major source of heat now centered under Yellowstone (where there is active volcanism in the form of some rising magma, thermal hotsprings and geysers such as Old Faithful). The track of Yellowstone extends across Idaho and into southern Oregon, with older and older volcanic rocks as one moves westward.

While the up-welling of a plume under oceanic lithosphere causes melting that gives rise to basalts, which is because mantle rock is involved. However, up-wellings under continents require a much greater magma pathlength through the continental lithosphere and crust. This causes basaltic melts of mantle material to be contaminated by the host rocks, enriching the magmas in silica and volatiles. Thus, some hotspot regions on continents have had massive explosive eruptions, unlike the effusive flows founds on Hawaii. The Yellowstone hotspot track has several calderas along its length, where past massive explosions have occurred, dwarfing even the vast explosion that produced the Long Valley Caldera of eastern California. Thus, hotspots do have some catastrophic potential, but it is fairly limited due to the small percentage of hotspot volcanoes on the surface (5%).

The island of Iceland is actually a hotspot volcano located right on the mid-ocean rift in the northern Atlantic. The combined volcanism of the hotspot and sea-floor spreading have built up the island to above sea-level. Iceland has extensive volcanic activity along the rift that bisects it from south to north, and is extensively studied as a natural laboratory for what happens at mid-ocean ridges in general. But, there is the unusual juxtaposition of the hotspot that complicates any generalizations based on Iceland observations. For the most part hotspot volcanoes are located within the interiors of plates, and appear to have fixed deep origins, below the plates. Most have effusive basaltic eruptions, but there are some on continents with explosion capabilities.

All volcanic eruptions, whether they involve arc, rift or hotspot volcanoes, clearly deliver material from the interior to the surface. The extruded rocks build up mountains, islands, and lava flows, some of which have added to the size of continents. The gases and volatiles brought up to the surface have built up the atmosphere and oceans. From the amount of certain elements in the current atmosphere, such as Neon, we are confident that the Earth that accreted had no initial atmosphere, which must have boiled off during the heavy bombardments and magma ocean phase. Thus, all of the present atmosphere and oceans have come from the interior (apart from minor additions from infalling comets). Looking at how volatiles come out of the ground today, we are convinced that volcanoes are now, and presumably have always been the main sources of gas transfer from the interior to the surface. The current atmosphere is 79% nitrogen and 21% oxygen, with traces of water and other materials. Nitrogen is being released at many volcanoes today, and fortunately the Earth is warm enough that nitrogen has not combined with hydrogen in the Earth's atmosphere to condense out as ammonia (as appears to be the case on the major gas-planets such as Jupiter). While free oxygen is not coming out of volcanoes, carbon dioxide does, and plant respiration is responsible for the conversion to an oxygen atmosphere, as we noted before.

We can test the gasses coming out of volcanoes today to see if it is consistent with what is found in the atmosphere. Volcanic gas samples reveal large amounts of water (H2O), carbon dioxide (CO2) and Nitrogen (N2) coming out today, along with smaller amounts of sulfur dioxide (SO2), Hydrogen sulfide (H2S), carbon monoxide (CO), Hydrogen (H2), hydrochloric acid (HCl), and Methane (CH4). If we take the current rate at which theses gases are emerging, and calculate the cumulative volumes over the history of the planet, we can only account for about 25% of the total water, chlorine, and nitrogen at the surface. This implies that if volcanism has always been the main source of atmospheric gases, there must have been more intensive volcanism, and perhaps larger gas fluxes in the past. Some materials, such as sulfur are less abundant on the surface than would be expected based on the sulfur expulsion rate. This implies that there is some mechanism for eliminating sulfur from the atmosphere. Sulfur is a very reactive material, and can combine with other elements to produce minerals that are efficiently subducted, thus cleansing the surface of much of the sulfur that emerges.

In addition to the gases that come out of volcanoes, there are large amounts of solid materials ejected, typically the smaller particles of dust reach the highest levels in the atmosphere. For strong vertical eruptions, dust can be propelled up into the stratosphere, above 17 kilometers. Once there, the suspended dust particles can block solar radiation, effectively heating up the stratosphere while the lower troposphere cools. This eruptive cooling was first noted by Ben Franklin in 1783 while he was in Europe, and he attributed the cooling to eruption of Laki on Iceland that year. The most important eruptions for influencing climate are massive vertical eruptions, which propel material into the stratosphere. Examples include the 1815 eruption of Tambora, which gave out 40 cubic kilometers of material and the 1883 eruption of Krakatoa. The eruption of Mt. Mazama which was the volcano preceding the collapsed caldera of Crater Lake, Oregon erupted in about 4000 B.C., sending out 5 cubic miles of material, more than 100 times that produced by the 1980 eruption of Mt. St. Helens.

Around the turn of the century, there was a spate of particularly large eruptions on Earth, with the 1883 eruption of Krakatoa, the 1902 eruptions of Pelee, Soufriere and Santa Maria, and the 1912 eruption of Katmai. Tracking of the average temperature around the Earth over the past 100 years, shows that this time was relatively cool, by about half of a degree compared to the period from 1920-1945 which had few eruptions. From 1945 on there have been more eruptions, which is slowly causing the temperature to decrease. The long-term oscillations of temperature associated with volcanism must be understood when investigating global warming or global cooling phenomena.

One of the most obvious aspects of volcanism is that it involves heat, and this energy source is inviting to tap. In the past few decades there have been numerous attempts to exploit Geothermal energy, in both volcanic and other areas. Several approaches have been tried, including hydrothermal energy, which involves bringing to the surface water heated by interaction with hot rocks at depth, and then using the steam energy for power. The U.S. potential energy from this source is twice the total energy in the world's oil supply. While the U.S. does not have that many volcanic areas with large hydrothermal systems, there is heat in all of the rock beneath our feet, and efforts to tap that heat are being explored. This U.S. abundance of so-called hot dry rock could provide 6000 times the world's supply of oil, if we can perfect technologies to extract the heat. Finally, magma reservoirs are sometimes shallow enough that we might be able to directly tap the heat in the molten rock. For the U.S. this would supply about 80 times the world's oil supply. So, with this bounty of ready energy below us, why do we still use fossil fuels? In part, it has proven difficult to efficiently extract much of the heat energy. For example, geothermal systems can quickly be drained of water, so that new water has to be input by recharge systems that require lots of energy themselves. While the water may heat up as it penetrates down to near a 900-1200 degree magma body, thus rising as steam with temperatures of 100-350 degrees C, the hot water is very reactive and often cements up the porous rock with minerals leached from the rock itself. Thus, there tends to be a finite lifetime to the circulation patterns in the ground, and it can be hard to sustain the productivity of the hydrothermal system.

Nonetheless, a few areas around the world are producing significant power from hydrothermal systems. Iceland is one of the most advanced nations in this respect, with 70% or so of the power for the capital city of Reykjavik being produced by hydrothermal power. The Geysers in northern California produces 500 megawatts of power per year, with 240 degree steam. This is enough power to meet the needs of San Francisco.

Another benefit of volcanic systems is that minerals and ores are concentrated near them. Water that emerges from or interacts with a magma body tends to be enriched in materials such as Fl, S, Zn, Cu, Pb, U, Au, Ag, Hg. These materials can precipitate out as the water circulates through the crust and cools in hydrothermal veins. This has given rise to major ore deposits. Some forms of volcanic up-wellings rise extremely rapidly through continental lithosphere, bringing high pressure minerals to the surface. This is how diamonds reach the surface. Diamond is a high pressure form of pure carbon, relatively unstable at the surface, where the common form of carbon called graphite is most stable. Diamonds are rapidly brought up from depths of 100-300 km in upwellings called Kimberlites. Thus, some of the glory of diamond rich rock in the interior are shared with us at the surface due to volcanism.

The societal response to natural hazards posed by earthquakes and volcanoes is couched in terms of the specific hazards that they present, as well as the viable options for dealing with the phenomena. We'll consider some of the specific hazards associated with each, and discuss the mitigation strategies that have evolved.

Earthquake hazards:

Primary: Fault Displacement/Ground Rupture

Secondary: Landslides

Earthquake Hazard Mitigation Approaches:

1. Construction Standards

Engineering Design

Building Codes

Retrofitting of Old Structures

2. Civil Preparation

Education

Emergency Response Planning

Fire Control

Disaster Response Teams

3. Hazard Analysis

Zoning

Dam/Nuclear Plant/Critical Facilities Planning

4. Timely Warning

Long-term Prediction

Short-term Prediction

Post event Anticipation

Key to making decisions regarding earthquake mitigation is use of the historical earthquake behavior and understanding of earthquake faulting driving mechanisms to develop seismic risk maps. These involve a consideration of the probability of certain levels of ground shaking during a given interval of time (often 30 or 50 years, a typical building life expectancy). Risk assessment also entails some consideration of the exposure to damage, which is influenced by population density, historical building practices, geography and other factors.

At some level, any attempt to define the probability of future ground shaking entails forecasting of future events. When seismologists discuss earthquake prediction, there is a very specific connotation of the term: an earthquake prediction must specify the location, magnitude and time of the event, given in terms of ranges or windows of parameters, with some assessment of the statistical likelihood of the event. This requires a very complete understanding of the physical processes producing an earthquake, and while there have been a few accurate 'predictions', based on various arguments, there is no accepted universal criteria for prediction specificity. But, we do have some general understanding of past earthquake behavior upon which we can base statistical projections of future events, albeit with great uncertainty. In this regard, earthquake prediction is akin to weather forecasting, entailing a future projection of the statistical likelihood of a particular outcome given a recent type of behavior, relative to past observations.

Earthquake forecasting is significantly more complex than weather forecasting, both because we have a less complete fundamental understanding of earthquakes and far fewer opportunities to measure the complexity of stress and deformation ongoing in the Earth relative to our ability to monitor global atmospheric conditions and variations of solar heating and other driving forces for weather.

We subdivide earthquake prediction efforts into

1. Long term forecasts

1. recurrence intervals

2. seismic gaps

2. Short term forecasts

1. precursors

2. post-event warning

Long term forecasts exploit our general understanding of plate tectonics, in that a given plate boundary fault is anticipated to have a cumulative faulting offset with time that corresponds to the relative plate motion. Each segment of the fault must slip a comparable total distance, even if some regions fail more often in smaller events while other regions fail rarely in large events. This notion has observational support in that earthquake rupture zones tend to fill in the entire plate boundary as a function of time, with rerupturing of the same segment being very common. For example, if we look at the Pacific/Eurasia plate boundary along Japan, where oceanic plate is underthrusting the islands, we find that in this century earthquakes have pretty uniformly covered the plate contact. In the previous century the same is true, with each segment of fault tending to rupture every 50-100 years on average. The slip in each event tends to be 5 m or so, which corresponds to the average time between ruptures multiplied by the plate convergence rate of about 9 cm/yr. In some regions we have over a thousand year history of repeated failure of a particular stretch of fault. Along the Tonankai-Nankaido region south of Honshu the average repeat time is about 180 years, although there are large fluctuations in the time between events. The Tokai Gap south of Tokyo is a region that failed in 1854, but did not fail when adjacent areas of the fault ruptured in 1944 and 1946.

A seismic gap is a region that is known to have previously failed in earthquakes (versus a continuously creeping section of fault), and has not failed for a length of time close to the average recurrence interval for that stretch of fault. Long-term forecasts try to identify which faults are 'mature' in terms of approaching their average recurrence delay since the last event. This notion assumes some uniformity in the rate of plate motions and the limiting strength of rock (bounding the strains that can accumulate prior to frictional sliding). Seismic gaps can be assigned probabilities for failure based on the spread in repeat times between earthquakes, although the latter is not well known for many faults, and for some faults has huge variance. Long-term forecasts based on recurrence behavior are useful for societal planning (construction codes, etc., acknowledging the regional hazard), but are not useful for evacuation or other short-term decisions.

Short-term prediction or forecasting is based on the notion that there will be some preparatory phenomena in the rock mass around a fault as it approaches the limiting strain conditions which induce frictional instability. The idea is that the instability must have some detectable precursory effects, since the highly strained rock is on the edge of an instability prior to sliding. One of the most commonly sought effects is precursory seismicity.

Seismicity patterns that are observed prior to some earthquakes include:

1. foreshocks (must be distinctive from the ensuing mainshock to be useful)

2. spatial-temporal migrations of seismicity

3. quiescence (a temporary shut-off of seismicity in a region prior to failure)

4. doughnuts (circular patterns of seismicity surrounding a patch that is about to fail)

5. swarms (bursts of numerous events with no clear mainshock).

Other effects that are sought include strain effects that result in anomalous behavior just before an earthquake. Some of the things that are measured are:

1. uplift/deformation, the problem is to detect anything unusual just before failure, not simply the steadily accumulating deformation

2. changes in P and S velocity of rock. The idea here is that cracks begin to open just before the rock slides, due to rock dilatancy (expansion) This can produce a temporal decrease in seismic wave velocities near the fault zone of the imminent earthquake. Not found to be very reliable.

3. groundwater variations. As cracks open, water can be flushed out and well levels can rise or fall anomalously. Hard to separate effect from normal fluctuations.

4. radon, He gas emissions. As cracks open the rate of gas flux from the ground may vary

5. electrical or magnetic properties. Changes in fluids in the rock may affect currents in the ground.

The hope is that the volume of rock that is undergoing final strain accumulation right up to the point of sliding failure is both large enough to result in a measurable surface effect, and that it will be temporally recognizable. We find that some events do seem to be preceded by one or more of these strain like effects, but other events have no precursor.

More exotic approaches include

1. animal behavior, presumably animals may sense tiny earthquakes, gas emissions, electrical currents etc., but probably instruments are more reliable. Unless we just don't know what to record.

2. tidal triggering/planetary alignments. The idea is that the gravitational pull on the Earth's crust from other bodies in the solar system may trigger faulting. There has been no convincing demonstration of such an effect related to lunar tides, which are far stronger than those from other planets or conjunctions of planets.

3. earthquake sounds and lights. There are clearly such phenomena during earthquakes, but no clear precursor has been documented.

A few prediction case histories

I. 1975 Haicheng earthquake. In this northern China city, a prediction was made based on anomalous foreshock earthquake activity in a normally quiet region. Well levels, tilting of the ground and anomalous animal behavior were observed, along with curious effects such as spontaneous ignition of gas bubbles in swamps.

The prediction evacuated the town of 100,000 for 2 nights, with a magnitude 7.3 earthquake occurring on the second day. 90% of the houses collapsed or were damaged, and few lives were lost. The evacuation was successful in part because of the totalitarian regime which enforced it, along with popular awareness of earthquake hazard and detection of the small foreshocks. It also was fortunate that the event happened within a short time, or else the evacuation may not have been sustained.

II. 1978 Oaxaca, Mexico. In 1977 U. Texas researchers observed temporal quiescence in a substantial seismic gap, which began in 1973. They estimated a magnitude 7.5 event could occur. In August 1978, small foreshocks were detected by a small seismic array that had been installed in the region. In November 1978, a magnitude 7.8 event occurred. This is an example of a medium-term forecast (no precise time was given for the event, although a magnitude and location were).

III. 1981 Peru. A major seismic gap along the coast has persisted for over 100 years since magnitude 8.0+ earthquakes occurred last century. U.S. Scientists Brady and Spence developed a complex theory of earthquake occurrence. They made very specific predictions based on their theory, for a magnitude 8.8 on August 10, 1981 and a 9.8 on Sept. 15, 1981. The U.S. National earthquake prediction council evaluated this prediction and found no credibility to the theory. Nonetheless there was great panic in Peru, with devastation of tourism, exodus of wealthy people and extension depression. The prediction was withdrawn as the specific occurrence of a small foreshock did not take place, but the damage was done. This was a major international fiasco.

IV. Parkfield, California. On a small stretch of the San Andreas fault in central California, a moderate size, magnitude 6.0 earthquake occurred in 1966. Investigations showed that the same stretch of fault had failed in 1934, 1922, 1901, 1881, and 1857. The times between events were then 32, 12, 21, 20, and 24 years, with a mean repeat time of 22 years and a standard deviation of 3 years. Based on this behavior, the U.S. Geological Survey made its only official prediction, for a magnitude 5.5 earthquake in 1988+/- 5 years. They set up a monitoring system to seek any short term precursors as well. At the end of 1993 the prediction window was exceeded and the prediction formally failed. It is now thought that there will be an earthquake in the future at Parkfield, but there appears to be more variability than the last 6 events had suggested. A concern is that the 1857 event occurred shortly before the 1857 rupture of the southern San Andreas fault in a great, magnitude 8 earthquake. Will the next Parkfield event trigger the Big One?

V. Tokai Japan. Based on the presence of a seismic gap south of Tokyo, which last ruptured in 1854, and before that in 1707, it is expected that there will be an event as large as 7.5-8.0. The adjacent area of fault ruptured in 1944. The Japanese have set up a massive earthquake prediction program, with extensive measurement of all viable phenomena which may show precursors (seismicity, radon gas, electrical, magnetic, tilting, water, etc.). Official procedures have been set up with a set of prediction criteria and an advisory panel that will make the prediction. Failure is not allowed, nor is uncertainty.

While specific predictions have a checkered history, there are now maps of probability of earthquake shaking for many regions of the San Andreas, based on historic activity. Southern California is deemed to be more likely to have the next major event than Northern California, but the Parkfield experience has increased skepticism.

In some cases, we can exploit local detection of an earthquake to phone ahead and warn that the seismic waves are on the way. This idea is used in Tsunami warning system for the Pacific. In this case, seismic recordings are used to locate events around the Pacific, to determine their size, depth, and faulting mechanism. If the event appears to be a good candidate for having displaced a lot of water, a tsunami warning is issued. This alerts areas far from the earthquake that a sea wave may be approaching. This is possible because the seismic waves travel much faster than tsunamis in the ocean do (they travel about the speed of a jet plane, so it takes 4-10 hours to traverse the Pacific). Events can be studied by seismic waves in a few tens of minutes to an hour. This allows alerts to be broadcast for places like Hawaii, Alaska, Japan, and other areas that are vulnerable to tsunamis from earthquakes at various positions around the Pacific. It is not useful for the very local tsunami, as that arrives before the seismic signals can be analyzed.

An analogous system was set up in 1989 after the Loma Prieta earthquake. The collapse of the Cyprus Freeway in Oakland required emergency crews to risk their lives going into the pancaked freeway. Aftershocks threatened to make further collapses. What was done was to set up a phone system, by which aftershocks that were felt strongly in San Jose led to alerts in Oakland via electronic transmission, far faster than the speed of seismic waves. Thus, there were a few tens of seconds for rescue workers to get out prior to the arrival of shaking from aftershocks 100 km to the south.

Aside from the scientific issues of earthquake prediction there are major socioeconomic considerations that influence the process. An earthquake prediction can cause great economic impact, decreasing property values, business activity, leading to civic disruption. Prediction assessment panels have been set up to try to maintain control on the process, but there is much concern about the liability incurred by a scientist who makes a prediction that fails. There are also dilemmas about whether a prediction is actually beneficial in some cases, as it may prompt response that exacerbates the losses even if the prediction is correct. For example, panic evacuations may clog freeways that collapse during an event or may prevent emergency vehicles from being able to respond to fires and injuries. It is no surprise that few people are bold enough to publicly make a prediction.

Some of the major volcanic hazards are:

1) Primary Lava

2) Secondary Lahars

Lava flows are usually not very dangerous, as they move rather slowly. But, there are exceptions. For example, the 1977 Nyiragongo (Zaire) volcano had a side fissure drain the lava lake in the main crater very rapidly, with lava squirting out at 60 miles per hour. This caused 1000 fatalities. Hawaii has had relatively rapid flows run into developed areas, but usually evacuation is possible.

Ash falls are typified by the 79 AD eruption of Mt. Vesuvius, which blanketed Pompeii and Herculaneum with mud ash and gas, causing about 20,000 fatalities. The main U.S. concern is over the downwind (easterly) deposit of ash from Cascade volcanoes, as occurred in 1980 for Mt. St. Helens.

Nuee Ardente are fast moving flows of gas and magma, in a volcanic avalanche. These are very deadly, as they flow fast and with utter destruction. Examples include the 1902 eruption of La Soufriere, St. Vincent, which killed 1500, and the 1902 eruption of Mt. Pelee, Martinique, when the town of St. Pierre was overrun, killing 30,000 (leaving only 2 survivors in jail). The latter event was particularly sad, as evacuation had been encouraged based on activity of the volcano and massive migrations of snakes through town away from the volcano.

Gas emissions accompany every volcano, but in some cases there are special conditions that allow the gasses to build up. The 1783 eruption of Mt. Laki, Iceland had a massive flux of sulfuric and fluorine gas, which killed great numbers of livestock. The ensuing famine led to 10,000 fatalities. The 1986 Lake Nios, Cameroon event involved an overturn of CO2 that had accumulated at the base of a lake in the crater. Over 1500 people were asphyxiated by the gas cloud that bubbled forth.

Lahars are volcanic mudflows, typically resulting from eruption under ice and snow at the summit, which mixes with the ash to make fast flowing muds. The 1985 Nevado del Ruiz event involved a 12 foot mudflow that descended on Armero, Colombia, killing 23,000. There had been six hours to evacuate, but this was not enough. Such mudflows occur again and again, in predictable paths, so some planning can be done. The main U.S. concern is for Mt. Rainier, where past mud flows run into Seattle.

Tsunamis occur when eruptions displace ocean water. The 1883 Krakatoa, Java eruption ejected 5 cubic miles of material, producing a 100 foot high wave which spread out and destroyed 300 towns, taking 36000 lives. The 1450 B.C. eruption of Santorini ejected almost 25 cubic miles of rock, producing the huge sea waves that swept away Minoan civilization.

Agricultural catastrophes are secondary consequences of the widespread devastation of volcanic eruptions. The 1815 Tambora eruption covered 1 million square miles with ash, with 40 cubic miles of volume. While this took 10,000 lives directly, about 80,000 perished due to famine. In 1902, Santa Maria volcano in Guatemala had a 5.5 cubic kilometer eruption. The ash killed many birds, leading to flies and mosquitoes flourishing. Malaria outbreaks ensued and are blamed for 3000 deaths.

These hazards are varied and clearly not easily controlled. Indeed it is a basic fact that we cannot do much to limit damage from volcanic hazards by better construction methods and the like. The emphasis is on prediction of the event so that evacuation can save lives.

Volcanic prediction is difficult, but in many ways it is more viable than for earthquakes. The major difficulty is that every volcano has distinctive behavior, which must be characterized case by case. This is challenging because of the long period of repose between explosions.

Most volcanic predictions are based on various phenomena:

1. Statistical behavior of that volcano

2. Earthquake activity, often tracking the ascent of magma

3. Ground deformation

4. Changes in temperature of the gases and lava pool

5. Changes in chemistry of the gas

For some volcanoes, the statistical behavior can be characterized if there is a history of 10-20 eruptive sequences. This is an empirical approach, as there is no simple physical theory for the eruptive cycle. Volcanoes are observed to have either random eruption sequences, or sequences in which the probability of an eruption increases with time since the last eruption, or in which the probability decreases with time since the last eruption. The latter case corresponds to bursts of activity followed by repose, while the middle case is gradual accumulation of pressure during periods of repose. Individual behavior may be stable over fairly short periods of time, even though it may change through the lifetime of the volcano.

Statistical methods give gross probabilities, but not short-time prediction. Earthquake activity near volcanoes is due to stresses in the crust, temperature changes, magma movements, and gas explosions. Earthquake swarms are often precursors to eruptions, especially when the depth of events decreases with time, reflecting ascent of magma. Prolonged ground vibrations called harmonic tremor are associated with resonances in the fluid filled plumbing of the volcano. On average when an increase in earthquake activity is observed, 58% of the time it precedes an eruption, 38% of the time there is no eruption, and 4% of the time there is an eruption with no earthquakes. Individual volcanoes differ in the reliability of earthquake precursors, and again each volcano must be characterized.

Inflation of the magma chamber below the volcano causes tilting and uplift of the surface which can be measured. This is the direct result of ascent of magma and build-up of gas pressure. For Kilauea, the probability of eruption increases with increasing tilt of the surface. But, the problem is often that tilting occurs, but magma may not reach the surface. How to anticipate actual outpourings is a problem.

There are many instances of successful prediction of major eruptions. We will see a case in the movie about Mt. Pinatubo. But, a sobering case history is offered by Soufriere of Guadelupe in 1976. This volcano had come under observation due to small steam explosions and earthquake tremors. Numerous scientific teams arrived, and bickered over what to make of the activity, and the media played up the arguments. Ultimately, a town of 74,000 was evacuated for 4 months, at a cost of $500 million dollars. But, no eruption happened. The agonizing over this choice may have played a conservative role in the reluctance to evacuate Armero in 1985 when the Nevado del Ruiz lahar destroyed the city and killed 23,000.

The U.S. has set up a hierarchy of volcano warning levels:

1. Notice of Potential Hazard

2. Hazard Watch - A indefinite state of observation

3. Hazard Warning - Specific time, location, explosion size given

The 1980 eruption of Mt. St. Helens came during a Hazard Warning based on seismic activity, inflation of the summit and history of the mountain, which has had many lateral explosions and violent eruptions.