ABSTRACT Kīlauea Volcano’s active summit lava lake posed hazards to downwind residents and over 1.6 million Hawai‘i Volcanoes National Park visitors each year during 2008–2018. The lava lake surface was dynamic; crustal plates separated by incandescent cracks moved across the lake as magma circulated below. We hypothesize that these dynamic thermal patterns were related to changes in other volcanic processes, such that sequences of thermal images may provide information about eruption parameters that are sometimes difficult to measure. The ability to learn about concurrent gas emissions and seismic activity from a remote thermal time-lapse camera would be beneficial when conditions are too hazardous for field measurements. We applied a machine learning algorithm called self-organizing maps (SOM) to thermal infrared time-lapse images of the lava lake collected hourly over 23 April–21 October 2013 ( n = 4354). The SOM algorithm can take thousands of seemingly different images, each representing the spatial distribution of relative temperature across the lava lake surface, and group them into clusters based on their similarities. We then related the resulting clusters to sulfur dioxide emissions and seismic tremor activity to characterizeties between the SOM classification and different emplacement conditions. The SOM classification results are highly sensitive to the normalization method applied to the input images. The standard pixel-by-pixel normalization method yields a cluster of images defined by the highest observed SO 2 emission levels, elevated surface temperatures, and a high proportion of cracks between crustal plates. When lava lake surface patterns are isolated by minimizing the effect of temperature variation between images, relationships with seismic tremor activity emerge, revealing an “intense spatter” cluster, characterized by unstable, broken-up crustal plate patterns on the lava lake surface. This proof of concept study provides a basis for extending the SOM classification method to hazard forecasting and real-time volcanic monitoring applications, as well as comparative studies at other lava lakes.

Abstract During the period 28 February 2000–31 December 2013, the MODVOLC system ( http://modis.higp.hawaii.edu ) autonomously analysed almost 9 trillion (i.e. 9×10 12 ) pixels contained within almost 3 million MODIS images, searching for evidence of high-temperature thermal signatures associated with volcanic eruptions. Thermal unrest, mainly associated with active lava, be it in the form of flows, domes, lakes or confined to vents, was detected at 93 volcanoes during this period of time. The first part of this paper describes the physical basis and operational implementation of the MODVOLC algorithm. The second part presents data to detail the nature of the thermal emission from these 93 volcanoes over the past 14 years.

Abstract Many volcanoes around the world are poorly monitored and new eruptions increase the need for rapid ground-based monitoring, which is not always available in a timely manner. Initial observations therefore are commonly provided by orbital remote sensing instruments at different temporal, spatial and wavelength scales. Even at well-monitored volcanoes, satellite data still play an important role. The ASTER (Advanced Spaceborne Thermal Emission Radiometer) orbital sensor provides moderately high spatial resolution images in multiple wavelength regions; however, because ASTER is a scheduled instrument, the data are not acquired over specific targets every orbit. Therefore, in an attempt to improve the temporal frequency of ASTER specifically for volcano observations and to have the images integrate synergistically with high temporal resolution data, the Urgent Request Protocol (URP) system was developed in 2004. Now integrated with both the AVHRR (Advanced Very High Resolution Radiometer) and MODIS (Moderate Resolution Imaging Spectroradiometer) hotspot monitoring programmes, the URP acquires an average of 24 volcanic datasets every month and planned improvements will allow this number to increase in the future. New URP data are sent directly to investigators responding to the ongoing eruption, and the large archive is also being used for retrospective science and operational studies for future instruments. The URP Program has been very successful over the past decade and will continue until at least 2017 or as long as the ASTER sensor is operational. Several volcanic science examples are given here that highlight the various stages of the URP development. However, not all are strictly focused on effusive eruptions. Rather, these examples were chosen to demonstrate the wide range of applications, as well as the general usefulness of the higher resolution, multispectral data of ASTER.

Abstract The Volcano Sensor Web (VSW) is a globe-spanning net of sensors and applications for detecting volcanic activity. Alerts from the VSW are used to trigger observations from space using the Earth Observing-1 ( EO-1 ) spacecraft. Onboard EO-1 is the Autonomous Sciencecraft Experiment (ASE) advanced autonomy software. Using ASE has streamlined spacecraft operations and has enabled the rapid delivery of high-level products to end-users. The entire process, from initial alert to product delivery, is autonomous. This facility is of great value as a rapid response is vital during a volcanic crisis. ASE consists of three parts: (1) Science Data Classifiers, which process EO-1 Hyperion data to identify anomalous thermal signals; (2) a Spacecraft Command Language; and (3) the Continuous Activity Scheduling Planning Execution and Replanning (CASPER) software that plans and replans activities, including downlinks, based on available resources and operational constraints. For each eruption detected, thermal emission maps and estimates of eruption parameters are posted to a website at the Jet Propulsion Laboratory, California Institute of Technology, in Pasadena, CA. Selected products are emailed to end-users. The VSW uses software agents to detect volcanic activity alerts generated from a wide variety of sources on the ground and in space, and can also be easily triggered manually.

Abstract Infrared (IR) satellite-based sensors allow the detection and quantification of volcanic hot spots. Sensors flown on geostationary satellites are particularly helpful in the early warning and continuous tracking of effusive activity. Development of operational monitoring and dissemination systems is essential to achieve the real-time ingestion and processing of IR data for a timely response during volcanic crises. HOTVOLC is a web-based satellite-data-driven monitoring system developed at the Observatoire de Physique du Globe de Clermont-Ferrand (Clermont-Ferrand), designed to achieve near-real-time monitoring of volcanic activity using on-site ingestion of geostationary satellite data (e.g. MSG-SEVIRI, MTSAT, GOES-Imager). Here we present the characteristics of the HOTVOLC system for the monitoring of effusive activity. The system comprises two acquisition stations and secure databases (i.e. mirrored archives). The detection of volcanic hot spots uses a contextual algorithm that is based on a modified form of the Normalized Thermal Index (NTI*) and VAST. Raster images and numerical data are available to open-access on a Web-GIS interface. Tests are carried out and presented here, particularly for the 12–13 January 2011 eruption of Mount Etna, to show the capability of the system to provide quantitative information such as lava volume and time-averaged discharge rate. Examples of operational application reveal the ability of the HOTVOLC system to provide timely thermal information about volcanic hot spot activity.

Abstract Effusion rate is a crucial parameter for the prediction of lava-flow advance and should be assessed in near real-time in order to better manage a volcanic crisis. Thermal remote sensing offers the most promising avenue to attain this goal. We present here a ‘dynamic’ thermal proxy based on laboratory experiments and on the physical framework of viscous gravity currents, which can be used to estimate the effusion rate from thermal remote sensing during an eruption. This proxy reproduces the first-order relationship between effusion rate measured in the field and associated powers radiated by basaltic lava flows. Laboratory experiments involving fluids with complex rheology and subject to solidification give additional insights into the dynamics of lava flows. The introduction of a time evolution of the supply rates during the experiments gives rise to a transient adjustment of the surface thermal signal that further compromises the simple proportionality between the thermal flux and the effusion rate. Based on the experimental results, we conclude that a thermal proxy can only yield a minimum and time-averaged estimate of the effusion rate.

Abstract We present a novel method for interpreting time series of multispectral observations of volcanic eruptions. We show how existing models relating to radiance and area emplacement can be generalized into an integration-convolution of a Net Area Emplacement (NAE) function and a cooling function, assuming all surfaces follow the same cooling curve. The NAE describes the variation in the rate of emplacement of hot material with time and temperature, while the cooling function describes the cooling of a hot surface with time. Discretizing the integration-convolution equation yields an underdetermined matrix equation that we solve using second-order Tikhonov regularization to stabilize the solution. We test the inversion by modelling plausible NAE surfaces, calculating the radiances, adding noise and inverting for the original surface. Three or more spectral bands are required to capture the overall shape of the NAE, and recovering specific quantities is difficult. Single wavebands that yield flat kernels recover the total area emplacement curve (rate of increase of hot area – the integral of the NAE with respect to temperature) surprisingly well due to their property of conserving NAE, suggesting novel methods for calculating area emplacement rates (and effusion rates) from time series of satellite images and radiometer measurements.

Abstract VolcFlow is a finite-difference Eulerian code based on the depth-averaged approach and developed for the simulation of isothermal geophysical flows. Its capability for reproducing lava flows is tested here for the first time. The field example chosen is the 2010 lava flow of Tungurahua volcano (Ecuador), the emplacement of which is tracked by projecting thermal images onto a georeferenced digital topography. Results show that, at least for this case study, the isothermal approach of VolcFlow is able to simulate the velocity of the lava through time, as well as the extent of the solidified lava. However, the good fit between the modelled and the natural flow may be explained by the short emplacement time ( c. 20 h) of a thick lava ( c. 5 m), which could limit the influence of cooling on the flow dynamics, thus favouring the use of an isothermal rheology.

Abstract Developments in spaceborne Earth Observation (EO) sensor technology over the last decade, combined with well-tested physical models and multispectral data-processing techniques developed from the early 1980s, have paved the way to the global monitoring of volcanoes by sensors of metric, decametric, kilometric and multi-kilometric spatial resolution. Such variable geometries provide for revisit intervals ranging from about monthly – at high-spatial resolution in Low-Earth Orbit – to less than 5 min – at low-spatial resolution, from geostationary platforms. There are currently about 20 spacecrafts available for carrying out 24/7 quantitative observations of volcanic unrest, at all resolutions and as close as possible to real-time. We show some successful examples of synergetic EO on volcanoes on three continents from 10 different payloads, automatically processed with three, end-to-end unsupervised procedures, on eight major eruptions and a lava lake between 2006 and 2014.

Abstract We describe a new volcanic hotspot detection system, named Middle InfraRed Observation of Volcanic Activity (MIROVA), based on the analysis of infrared data acquired by the Moderate Resolution Imaging Spectroradiometer sensor (MODIS). MIROVA uses the middle infrared radiation (MIR), measured by MODIS, in order to detect and measure the heat radiation deriving from volcanic activity. The algorithm combines spectral and spatial principles, allowing the detection of heat sources from 1 megawatt (MW) to more than 10 gigawatt (GW). This provides a unique opportunity to: (i) recognize small-scale variations in thermal output that may precede the onset of effusive activity; (ii) track the advance of large lava flows; (iii) estimate lava discharge rates; (iv) identify distinct effusive trends; and, lastly, (v) follow the cooling process of voluminous lava bodies for several months. Here we show the results obtained from data sets spanning 14 years recorded at the Stromboli and Mt Etna volcanoes, Italy, and we investigate the above aspects at these two persistently active volcanoes. Finally, we describe how the algorithm has been implemented within an operational near-real-time processing chain that enables the MIROVA system to provide data and infrared maps within 1–4 h of the satellite overpass.

Abstract Timely detection and quantification of lava effusion rates are crucial for volcanic hazard mitigation during effusive eruptions. Satellite-based detection methods typically exploit the exceptional radiant heat fluxes associated with lava effusion, but effusive eruptions can also emit prodigious amounts of sulphur dioxide (SO 2 ). Measuring the magnitude and temporal evolution of SO 2 emissions provides an additional means for monitoring effusive eruptions, complementing thermal monitoring. Examples of effusive eruptions detected since 1978 using ultraviolet (UV) satellite measurements of SO 2 emissions by the Total Ozone Mapping Spectrometer (TOMS), Ozone Monitoring Instrument (OMI) and Ozone Mapping and Profiler Suite (OMPS) are reviewed. During many effusive eruptions, trends in SO 2 production mimic the classic waxing–waning pattern characteristic of such events that is also seen in thermal infrared (TIR) hotspot data, suggesting a qualitative link between SO 2 emissions and lava effusion rates. An example of lava effusion rate calculation based on TOMS SO 2 measurements is presented for the 1998 eruption of Cerro Azul (Galápagos Islands), for which detailed eruption observations and independent estimates of effusion rates are available. Combining TOMS-derived SO 2 emission rates with estimates of sulphur content in Cerro Azul lavas yields lava effusion rates almost identical to independently derived values, demonstrating the utility of the technique.

Abstract Accurate and fast delivery of information about recent lava flows is important for near-real-time monitoring of eruptions. Here, we have characterized the October 2010 lava flow at Piton de la Fournaise using various InSAR datasets. We first produced a map of the area covered by the lava flow (i.e. Area lava =0.71–0.75 km 2 ) using the coherence of two syn-eruptive interferograms. Then we analysed two post-eruptive InSAR datasets (i.e. monostatic and bistatic data). The monostatic database provided us simultaneously with the displacement rates, lava thickness, volume and volume flux. We found that the lava flow was subsiding and moving eastward at maximum rates of 13±0.3 and 4±0.2 cm a −1 , respectively. Also, it had a mean thickness of Z mean =5.85 m, Vol DRE =1.77±0.75×10 6 m 3 (1σ) and MOR=1.25±0.53 m 3 s −1 . The bistatic database provided us only with the thickness and volume information (i.e. Z mean =6.00 m, Vol DRE =1.83±0.65×10 6 m 3 and MOR=1.29±0.46 m 3 s −1 ). Finally, we used a thermal remote sensing technique to verify the InSAR-derived measurements. Results show that the monostatic and bistatic datasets were both well within the range for the DRE volume obtained from MODIS data (2.44–4.40×10 6 m 3 ). Supplementary material: Tables A1 and A2 give satellite images used in this study. Table A3 gives the parameters used for the calculation of the effusion rates. The figures give the data processing of the post-eruptive radar images. These are available at https://doi.org/10.6084/m9.figshare.c.2213563