Gold Open Access: This paper is published under the terms of the CC-BY license.


We investigated the creation of a volcanic islet and emplacement of lava flows in the sea by analyzing data from the island-forming eruption at Nishinoshima, Japan, that has been continuing since November 2013. Aerial observations and satellite images were used to perform a quantitative analysis of the eruption processes. The most intriguing characteristic of the lava flows is the development of lobes and tubes from breakouts and bifurcations of andesitic ‘a’ā-type lava flows. Internal pathways that fed lava to the active flow front were eventually developed by crust solidification and dominated the lava transport. The average discharge was ∼2 × 105 m3/day, and the total volume of erupted material reached ∼0.1 km3 at the end of February 2015. Fractal analysis of the lava-flow margins suggests that the growth pattern is self-similar, with a fractal dimension (D) of ∼1.08–1.18, which is within the range of subaerial basaltic lava flows. The morphological evolution of Nishinoshima is controlled primarily by effusion of lava with an apparent viscosity of 104–106 Pa·s, average discharge of ∼2.3 m3/s, and eruption duration lasting ∼2 yr. Our data and analyses suggest that the effect of lava coming in contact with seawater, as well as the variations in the lava discharge rate on local and overall scales, are important factors affecting the development of crust and the lava transport system.


Lava-flow morphology reflects the emplacement and cooling dynamics of erupted lava, and is used to constrain eruption parameters, ambient conditions, and the physical properties of lavas (e.g., Huppert et al., 1982; Griffiths, 2000). Subaerial lava flows and their eruption processes are often recorded by monitoring instruments, and eruption parameters such as the discharge and its temporal variation are measured quantitatively (e.g., Harris et al., 2007). In contrast, lava flows in submarine or marine settings are rarely observed. Historical examples of eruptions in the sea that enable us to analyze time-series data and estimate eruption parameters are limited (e.g., Thordarson and Sigmarsson, 2009); therefore, the emplacement dynamics of lava flows in the sea are not well constrained. Since November 2013, the latest island-forming eruption comprising lava effusion and Strombolian activity has been occurring at Nishinoshima volcano, 130 km west of Chichijima, Ogasawara Islands, Japan (Fig. 1A). The lava flows have continued to reclaim land from the shallow sea. The new island merged with a pre-existing island and is still growing at present (October 2015). In this paper, we describe the eruption process and the morphological evolution of the lava flows that form a new volcanic islet, and estimate the eruption parameters, based on our aerial surveys and on satellite images taken by the TerraSAR-X and Pléiades satellites, both of which are operated by Airbus Defense and Space. TerraSAR-X offers synthetic aperture radar (SAR) images with a ground range resolution of 1 m, taken at intervals of 11 or 22 days. Pléiades produces 0.5-m-resolution optical color images. We also used image data with multispectral bands of 30 m and thermal bands of 100 m, taken by the Landsat-8 satellite operated by NASA and the U.S. Geological Survey (USGS). Results of theoretical and laboratory studies on compound lava flows (Blake and Bruno, 2000; Anderson et al., 2005) are applied to investigate the relationship between the morphological evolution of the new islet and the eruption parameters.


The 2013 eruption began with Surtseyan-type eruptions and the formation of a cone in a shallow sea of ∼20 m depth, ∼400 m southeast of the existing Nishinoshima Island. The exact date of the initial eruption is unknown, but a thermal anomaly was first detected in this area in satellite images from the Moderate Resolution Imaging Spectroradiometer (MODIS) in early November 2013. When a small islet was discovered by the Japan Maritime Self-Defense Force on 20 November 2013, it was 150 × 80 m in size. The new islet was also detected by TerraSAR-X on 22 November (Fig. 1B). On 21 November 2013, we carried out aircraft observations of this area and confirmed the Surtseyan eruptions. Within three days the eruption style changed to Strombolian, because a pyroclastic cone formed around the vent and prevented external water from flowing into the crater. During the Strombolian activity, an ‘a’ā-type lava flow emerged from the crater (Fig. 2). As the cone height increased, lava flows erupted from the crater or from openings on the flank of the cone. These flows branched, producing multiple lobes that extended the foreshore with newly formed land reclaimed from the shallow sea. In late December 2013, the new islet coalesced with the existing Nishinoshima (Fig. 1C). Strombolian activity and lava effusion continued for more than 16 months, further extending the new islet, with the lavas producing a substantial lava-flow field (Figs. 1D and 2). During the 16-month period, the surface area of the lava field increased by ∼2.6 × 106 m2 (at an average growth rate of ∼5.5 × 103 m2/day). The active crater has remained almost fixed at the center of the new island. The vent is in the same location as that recorded for one of the vents in the last eruption of A.D. 1973–1974. The apparently continuous magma supply with Strombolian eruptions and lava flows suggests that the magma supply from the chamber is steady. Government authorities have limited the access to the new islet, so we cannot analyze the new lava on site; however, the chemical composition of scoria and ash samples from the current Strombolian activity was obtained. The magma is a two-pyroxene andesite with a whole-rock composition of 58–59 wt% SiO2 (Saito et al., 2014), which is almost the same as in samples from the eruption of 1973–1974 (Umino and Nakano, 2007). The groundmass glass contains 63–65 wt% SiO2 and phenocryst content is ∼10 vol% (Appendix DR1 in the GSA Data Repository1).


One of the most intriguing characteristics of the growing islet is the multiply branching flow lobes that form numerous smaller lobes, each defined as tongue shaped and up to 100 m (typically 30–50 m) wide. Analyses of satellite images along with aerial observations (Figs. 1D and 2) show that the lava flows form a collection of overlapping and anastomosing lava lobes. Many of the lobes overrun each other. The observations also suggest that breakouts and bifurcations occur at the flow front when the lobes reach flat-lying ground with few topographic obstacles. Repeated breakouts have resulted in the formation of dozens of lobes. The development of the surface architecture that dominates the growth of Nishinoshima shows a feature consistent with “compound lava flows”, as opposed to simple lava flows (Walker, 1971; Kilburn and Lopes, 1988).

Based on aerial observations, we determined that the Nishinoshima lava flow is an ‘a’ā-type flow (Fig. 2). Levees alongside open lava channels were observed on the active lava lobes, becoming indistinct with time, and eventually being covered by a cool rubbly lava surface. These types of changes are clearly observed along the coast by comparison of lava lobes featuring a steam-generating front with those without steam (Fig. 2C). Clefts commonly develop at the lobe front and along the axial crest of lobes without levees. Analysis of a time series of satellite images shows that the clefts gradually open and widen with time (Fig. DR1 in the Data Repository). On the northern part of the island, 30-m-wide clefts developed over three months from January 2014. The cleft opening is occurring at an average speed of ∼0.3 m/day. The widening of the clefts correlates with an increase in the height of the lava field. The Geospatial Information Authority of Japan (GSI) (GSI, 2015) measured the topography of the island using an unmanned aerial vehicle. The data show that the height of the entire lava field increased with time, and the area with clefts thickened more than the rest of the flow field. These observations suggest that clefts are produced in the solidifying lava surface by inflation driven by an increase of internal pressure by successive injection of new lava into the lobes. Such lava-inflation clefts were documented in subaerial basaltic lava flows in Hawaii (i.e., Walker, 1991; Hon et al., 1994). Night aerial observations of Nishinoshima revealed glowing lava flow fronts (Fig. 2E). Landsat-8 thermal infrared images also show that the front of actively moving lava lobes had higher temperatures than other parts of the lava field (Fig. DR2). These observations suggest that molten lava is transported via tubes beneath the solidified crust to the flow front. This also demonstrates that the transport system of lava from the vent to the active flow front changes with time from open (lava channel) to closed (internal pathway) because of the solidification and thickening of the flow-top rubble and coherent crust. The development of this lava-field architecture is comparable to the formation of internal pathways within pāhoehoe sheet lobes (e.g., Thordarson and Sigmarsson, 2009).

Breakouts occur when the internal pressure in the lobe overcomes the strength of the crust (Walker, 1991). Breakouts develop when the solidified crust ruptures and new lava emerges from an earlier-formed lobe, spawning further breakout. The breakouts are divided into two types: (1) breakouts from the sides of the mother lobe producing lobes tens of meters wide, as seen in the satellite images (Fig. DR1), a type of breakout that frequently took place near the flow front; and (2) breakouts that produce larger lobes (hundreds of meters wide) originating at the central cone and subsequently covering large areas of the earlier lobes. These large breakouts occurred only three times (in late May to early June, early August, and mid-September 2014) but contributed significantly to the enlargement of the new island (Fig. 3). We infer that the large breakouts are linked to increases in the magma discharge as demonstrated below. Based on the TerraSAR-X images, a large breakout occurred on the eastern side of the island in late May 2014. At this time, it appeared that lava inflation on the western side ceased and along certain segments the solidified lava surface deflated. This observation suggests that the lava lobes are interconnected via an anastomosing transport system beneath the solid surface.


The variations in area, volume, and discharge rate were estimated based on the changes in the outline of the new islet obtained from TerraSAR-X, aerial observations, topography data of the new islet (GSI, 2015), and pre-eruption bathymetry data (Japan Coast Guard, 1993) (Appendix DR2). The volume above sea level (Vland) was calculated using the area, average thickness of the lava field, and the scoria cone shape estimate based on TerraSAR-X data and aerial observations. GSI topography data were useful to confirm the volume estimates and errors, but the measurements by GSI were infrequent (six times in a year). Therefore, we primarily used TerraSAR-X data (a total of 28 images) and compared these data to the GSI data, then estimated the variation of Vland. The volume below sea level (Vsea) was calculated by comparing the extent of the new islet with pre-eruption bathymetry data. The average discharge was calculated from the change in the total volume (Vland + Vsea) every 11 or 22 days, wherein a 15%–20% error derived from the uncertainty of the area estimation using TerraSAR-X images was considered.

The area of the new island grew at a relatively constant rate until May 2014 (Fig. DR6), but after that the growth was stepwise corresponding to episodic growth of the lava field caused by large-scale breakouts. The total volume of erupted material in the first 15-month period was calculated as ∼0.1 km3 (of which ∼60% was subaerial and ∼40% subaqueous). The discharge gradually increased to 2.5 × 105 m3/day in the first three months and declined in the following three months (Fig. 3B). Thereafter, the discharge fluctuated with a periodicity of about two months. When the discharge peaked during certain periods (labeled I, II, and III in Fig. 3B), a large-scale breakout was observed. The peaks correspond to the large breakouts that extended the flow-field area significantly (I, mid-May to June 2014; II, July to August 2014; and III, September 2014) (Fig. 3). The maximum discharge was 5.0 × 105 m3/day in September 2014, and the largest breakout occurred from the foot of the central cone on the north side at that time. The average discharge during the 15-month period was ∼2 × 105 m3/day (2.3 m3/s), which is similar to the discharge (2–5 m3/s) in the latest phase of the Surtsey eruption in A.D. 1964, although the discharge of the Surtsey eruption reached 20 m3/s on average in the early stages of effusive activity (Thordarson and Sigmarsson, 2009).

The fractal dimension of the lava-flow margin (D) is one of the quantifiable characteristics of lava-flow morphology and is used to evaluate the effect of rheology on the emplacement dynamics of lava flows (e.g., Bruno et al., 1994). We utilized a box-counting method to calculate D for the outlines of the Nishinoshima lava flows that spread along the shallow seafloor without a significant slope, using TerraSAR-X images (Appendix DR3). The results suggest that the lava flows are fractal, are controlled by a low-viscosity nature, and that their growth pattern is self-similar with D of 1.08–1.18 (Fig. 3C); these D values span the range of the basaltic ‘a’ā-type flows (1.05–1.09) and pāhoehoe-type flows (1.13–1.23) (Bruno et al., 1994).


Compound lava flows are commonly interpreted to be the products of low discharge during eruptions of low-viscosity magmas of basaltic composition (e.g., Walker, 1971; Kilburn and Lopes, 1988). The Nishinoshima lavas are andesitic; thus, the compound nature of the lava field would not be expected. Here, the relationship between the physical parameters and the morphology of compound lava flows (Blake and Bruno, 2000; Anderson et al., 2005) can help in understanding this phenomenon. Blake and Bruno (2000) proposed that a compound lava flow forms when the eruption continues over a period in which the effect of crust formation, hence the effect of the cooling rate, becomes important. The criterion is deduced from experimental observations and the theory of spreading viscous fluids, and is expressed as ηa < A(V/Q2), where ηa is the apparent viscosity (Pa·s), V is the volume (m3), and Q is the discharge rate (m3/s). A is a constant expressed as σc2s2κ/gρ (m2kg/s3), where σc is the crust strength at the flow margin (Pa), g is gravitational acceleration (m/s2), ρ is the density of the lava (kg/m3), κ is the lava’s thermal diffusivity (m2/s), and s is a non-dimensional coefficient. The cooling of the lava surface leads to formation of a thermal and hence rheological boundary layer of thickness δ, with , where t is the time; thus, the parameter s represents the cooling efficiency that is dependent on the thermal boundary conditions, heat transfer mechanisms, and the exact nature of the lava-flow front. Blake and Bruno (2000) used A = 0.002 m2kg/s3, assuming values found for subaerial Hawaiian lavas with a stationary crust (Hon et al., 1994). However, if the outer crust is continuously disrupted due to high discharge, exposing the incandescent interior of the lava, then the cooling rate will increase. Water is a more efficient coolant than air, and will therefore promote higher cooling rates. Simple heat transfer calculations lead to the deduction that, when lava flows contact water, s is expected to increase by one order of magnitude (e.g., Griffiths and Fink, 1992); therefore, in this case, A may increase by two orders of magnitude (i.e., A ≈ 0.2 m2kg/s3). A high value of V/Q2, or a high A, indicates the rapid cooling of the flow margin relative to the flow rate, leading to the emergence of many breakouts.

For Nishinoshima, using the volume and discharge data, V/Q2 over the eruption period (first 15 months) is calculated to be 106.8 ± 0.4 s2/m3. Thus, the apparent viscosity required to generate compound flows (ηa) can be constrained to <106 Pa·s using V/Q2 and the above criterion considering the enhanced cooling effect (Fig. 4). The viscosity is also constrained from petrological data (Appendix DR1). Using modal and chemical compositions of eruptive material from the current eruption, the relationship between the apparent viscosity (ηa) and the temperature (1060–1090 °C using pyroxene thermometers) can be calculated based on the method of Giordano et al. (2008) and Marsh (1981). The calculation results constrain the range of ηa for Nishinoshima magma to >104 Pa·s. Hence, the viscosity estimated from the theoretical criteria for compound lava flows is supported by petrological data, although it is relatively higher than that of known examples of compound lava flows with ηa of the order of 102–103 Pa·s and V/Q2 of 106–107 s2/m3 (e.g., Stasiuk et al., 1993) (Fig. 4). Our results suggest that in addition to an eruption condition with high V/Q2, the development of crust at the flow margins due to high cooling efficiency is an important factor that controls the breakouts and spreading of lava flows. The fractal nature of the lava-flow margins can be explained by the emergence of many breakouts caused by efficient cooling of low-viscosity lava. In fact, the development of the crust will be affected not only by the contact with water, but also by the variation of the discharge rate. A relatively low discharge rate promotes the formation of insulated transport and growth by lobe-to-lobe emplacement, while a relatively high discharge as well as conditions with high fluxes on a local scale promote the formation of open channels, resulting in ‘a’ā flows. Thus, the overall architecture of the Nishinoshima lava field may be controlled by the formation of crust from lava that has come in contact with seawater and by fluctuations in the overall magma discharge and/or local variations in the lava flux.


Aerial observations and satellite images were used to analyze the eruption processes at Nishinoshima. Lobes and tubes generated by andesitic ‘a’ā lava flows characterize the surface architecture of the lava field. Internal pathways that feed lava to the active flow front were eventually developed by crust solidification and dominated the lava-flow transport system. The morphological evolution of lava flows at Nishinoshima is controlled primarily by lava effusion with a viscosity of 104–106 Pa·s, discharge of ∼2 × 105 m3/day, and an eruption duration of ∼2 yr. The effects of the lava’s contact with seawater as well as the variations in the lava discharge rate on the local and overall scales are also important factors in the development of the crust and the transport system of the lava flows. The development of crust at the flow margins may be pronounced in eruptions in marine and lacustrine settings such as at Nishinoshima, and may be an enhancing factor in the emplacement processes of andesitic lava flows with multiple lobes, inflation clefts, and breakouts (e.g., lava flows from the Sakurajima volcano, Japan).

We thank PASCO Co. Ltd. for providing the satellite images taken by TerraSAR-X and Pléiades, both of which are operated by Airbus Defense and Space (, and acknowledge the Japan Coast Guard and the Geospatial Information Authority of Japan for providing the pre-eruption bathymetry data and topography data. The bathymetry data (M7023 ver. 2.0, Ogasawara area) issued by the Japan Hydrographic Association were used to produce Figure 1A. We also thank the crew and scientists on the R/V Natsushima who helped collect the ash samples on the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) NT15-E02 cruise. Our aerial inspections were carried out courtesy of the Asahi Shimbun and Mainichi Shimbun who kindly provided their aircrafts. We are very grateful to T. Thordarson, T. Gregg, and S. Rocchi for constructive comments.

1GSA Data Repository item 2016079, supplemental satellite images and petrological data, and volume estimation and image processing procedures, is available online at, or on request from or Documents Secretary, GSA, P.O. Box 9140, Boulder, CO 80301, USA.