In order to find an empirical event branch logic from abnormal phenomena to following volcanic activity for forecasting scale and type of eruption, the magma intrusion rate prior to eruptions of Sakurajima volcano is examined using ground deformation mostly from observation data and partially based on legends, for eruptions after the 20^th century: the 1914 eruption starting with plinian eruption followed by effusion of lava, the 1946 eruptions with lava effusion, eruptions at the summit crater of Minamidake during the period from 1955 to 2005, and vulcanian eruptions at Showa crater east of the summit from 2006 to 2017. Prior to the 1914 eruption, it is estimated that the magma intrusion rate attained a level of approximately 10⁸m³/day and was on the order of 10⁶m³/day during the effusion of lava in the 1946 eruption. During the eruptive period of Minamidake summit crater, three types of eruption occurred: vulcanian eruption, strombolian/lava fountain and continuous emission of volcanic ash. In cases of intrusion of magma forming a new conduit, the intrusion rate immediately before the 1914 eruption exceeded 10⁸m³/day, but only 10⁶m³/day in the dyke-forming event of August 15, 2015. Magma intrusion rate into a pre-existing conduit prior to eruptions at Minamidake summit crater are ordered as follows: vulcanian eruption (1×10⁵ to 8×10⁵m³/day)＞continuous emission of volcanic ash (approximately 1×10⁵m³/day)＞strombolian/lava fountain (0.2×10⁵ to 2×10⁵m³/day). The magma intrusion rate prior to vulcanian eruptions at Showa crater is smaller (approximately 10⁴m³/day) than for eruptions at Minamidake summit crater. However, the rate reached an order of 10⁵m³/day prior to lava fountain on August 22, 2017. Magma intrusion rates well correspond to the scale and type of eruption. In the case of magma intrusion under detection, the change of volcanic gas and increase in the heat discharge rate are available for the empirical event branch logic.

Eruptive activities of volcanoes are examined by using a global data catalog of volcanoes provided from Smithsonian Institution. Yearly numbers of volcanic eruptions with a volcanic explosivity index (VEI) larger than or equal to 2 are almost constant from 1900, and the frequency distribution of the magnitude of volcanic eruptions obeys a power law in the range of VEI≥2. About 10 to 30% of volcanic eruptions with VEI≤2 end within one day, and 8 to 15% continue for more than 1 year. On the other hand, a few percentages of large eruptions with VEI≥3 end for a few days, and 25-30% continue for more than 1 year. Once an eruption occurs, the occurrence possibility of new eruptions at volcanoes locating within a 100km distance increases about 30% for 0.2 year. When the volcanoes locating at a region with a radius of 200km are examined together, yearly numbers of eruptions per volcano fluctuate within a factor of three for a few tens of years at more than about 90% of the regions. Frequency distribution of the yearly number of eruptions follows an exponential decay, which suggests an existence of a characteristic frequency of the eruptive activity on the globe. These averaged features of eruptive activities at volcanoes around the world can be used to give some basic characteristics of the occurrence of new eruptions.

Ground deformation is often observed to precede eruptions. For assessments of volcanic activity and hazards, it is therefore important to clarify the relationship between the occurrence of ground deformation and any forthcoming eruptions. Here, we report volcanic ground deformation events that were detected at 9 volcanoes in Japan by applying the stacking method to strain measurements along baselines of GEONET, the GNSS network managed by the Geospatial Information Authority of Japan. We compiled the durations and amounts of strain change during 38 ground deformation events caused by deep pressure sources beneath the volcanoes. Sixteen of these 38 events were accompanied or followed by eruptions, whereas no deformation event was detected before or during seven eruption events. We compared the durations of ground deformation events to the elapse times between the beginning of deformation and eruption. In all ground deformation events preceding eruptions, the first eruption occurred during the ground deformation event. Eruptive activity ceased during the ground deformation event in 10 of the 16 eruptions preceded by deformation, and within the three months following the ground deformation event in 14 of the 16 events. The durations of ground deformation events not accompanied by eruptions (“failed eruptions”) were shorter than those of events accompanied by eruptions (“succeeded eruptions”). “Missed eruptions” were those not accompanied by deformation events. Succeeded eruptions accounted for 8% of the deformation events shorter than 180 days, and 60% of those longer than 180 days. The threat score is an evaluation of forecast accuracy on a scale from 0 to 1, with higher values indicating more accurate forecasts. Considering deformation events as eruption forecasts, we obtained a threat score of 0.36 for 44 events (including succeeded, failed, and missed eruptions, but excluding one difficult-to-classify event at Miyakejima), and a miss rate of 0.30. Importantly, the missed eruptions included two fatal eruptions. It is therefore difficult to use only this method to assess forthcoming eruptions. To improve the accuracy of eruption forecasts, it might be more efficient to evaluate ground deformation caused by shallow pressure sources.

Gabbroic inclusions newly found in the Yakeno lava flow, erupted about 1300 years ago on the northwestern flank of Fuji volcano, are divided into two groups; one consists of troctolite and the other consists of olivine gabbronorite and gabbronorite. Troctolite fragments are considered to originate from shallow dykes because of their mineral composition and high porosity. On the other hand, estimated equilibrium conditions of gabbronorite and olivine gabbronorite are 1020-1050°C and 250-380MPa, showing their deep origin; probably they had been caught in the Yakeno magma (i.e. the magma erupted as the Yakeno lava flow) at differentiated small magma bodies just above a substantial basaltic magma chamber of Fuji volcano. Pre-eruptive temperature and water content of the Yakeno magma are evaluated by olivine-liquid thermometer and plagioclase-liquid hygrometer to be 1090°C and 2.8wt%, respectively. So the viscosity of the Yakeno magma is calculated to be several tens Pascal second. Therefore the minimum ascent velocity required to lift the gabbroic fragments is ca.10m/h, implying that the Yakeno lava flow had ascent without significant pause en route to the surface.

We collected information on non-magmatic unrest events reported mainly of Japanese volcanoes that have been monitored for extended time periods, with the aim of conducting a statistical evaluation of these phenomena, which can sometimes lead to an eruption. We assessed the geomagnetic field changes and ground deformation, which indicate demagnetization and inflation, respectively. We then compiled the source depth, intensity, and the rate (i.e., magnetic moment rate / deformation rate), based on the single magnetic dipole model or the Mogi model, since these models are the most frequently cited in the literature. We found clear positive correlations between the source depth and its intensity and rate, exhibiting a linear trend on the log-log scatter plots. We confirmed that all the magnetic events and some of the deformation events were plotted markedly above the cutoff line related to the detection limit. We suspected that the geometry of the monitoring network and/or the simplistic point-source modeling could contribute to these positive correlations; otherwise some physical requirements such as subsurface structure might be the cause.

Although the number of unrest events collected was small, we investigated the validity of the evaluation of these unrest events from different viewpoints. Upward deviation from the linear trend on the scatter plot was found to be a potentially useful criterion, whereas the source depth had no definitive relationship with the subsequent occurrence of eruptions.

We assert that observable non-magmatic unrest events should be plotted within a diagonal area between the lower-right and upper-left domains in a scatter plot of source intensity (or rate) and depth. This is because weak sources at great depths cannot be detected, while overly strong sources close to the ground surface are physically impossible. In the intermediate zone, “unerupted” and “erupted” unrest events both occur. Based on this idea, we propose a statistical evaluation of the severity of unrest events, by measuring the deviation from the “unerupted unrests”, although this hypothesis must be tested with a sufficient number of unrest events in the future. Meanwhile, care must be taken in interpreting the data collected here, since they are based on the simplified point-source models that assume a uniform half-space, and they may not accurately recreate the physical entities.

Further studies that can relate the physical conditions and their time evolution in the subsurface to geophysical monitoring records at the ground surface will be necessary to evaluate how such unrest events are related to eruptions.

Volcanic gases are derived from a magma or a hydrothermal system, which are sources of magmatic or phreatic eruptions, respectively. Changes in a magma or a hydrothermal system, leading to eruptions, likely cause changes in volcanic gas emission rate and compositions. SO₂ emission rates correlate well with volcanic activities and precursory changes of the emission rates are also common. The SO₂ emission rate also tell us a volume and a degassing rate of the SO₂ sources, that can constrain processes of magma ascent or crystallization. CO₂/SO₂ ratios increases are observed prior to major eruptions of volcanoes with Strombolian activity by automatic monitoring of volcanic gas compositions, and are attributed to accumulation of CO₂-rich bubbles in a magma chamber. Common occurrence of excess degassing by explosive eruptions implies an eruption of a bubble-rich magma, suggesting that a bubble accumulation is a necessary condition of an eruption. Various precursory signals are observed prior to phreatic eruptions but quantitative modeling of the processes is not simple because the volcanic gas compositions are largely affected by various reactions at a shallow hydrothermal system. Combination of volcanic gas emission rate and composition monitoring data may help us for the quantitative evaluation of the volcanic gas changes associated with phreatic eruptions.

Based on the history of volcanic activity of Zao stage VI, we examined possible courses of future activity of Zao volcano. All activities will start with precursory phenomena. Next, phreatic eruptions from Okama crater or other place inner part of Umanose caldera will occur, and may cause ballistic materials release, ash fall, pyroclastic surge, and lahar. The possibility of small scales edifice collapse and lava flow swelled out is very low but should be included. When the activity progresses, magmatic eruptions will be taken place from Goshikidake, and cause same phenomena as in the phreatic eruptions but larger in scale. The possibility magmatic eruption takes place without preceding phreatic eruption can not be excluded. Rarely, the activity will go up further and resulted in sub-Plinian eruption. Aside from the above sequence, larger scale phreatic eruption from Goshikidake area should be listed, although the possibility of this is very low.