Analysis of thermal anomalies at Copahue Volcano between October 2011 and the December 2012 eruption with MODIS

Highlights

• Thermal anomalies recorded by MODVOLC at Copahue between 2000 and 2016 with NTI larger than −0.8 were co- and post-eruptive.

• We build time series of NTI and the MODIS radiance at 4 μm, and spatial and temporal evolution of the December 2012 eruption.

• An adapted threshold of −0.875 on the NTI enabled an appraisal of the genesis and evolution of the processes and emitted volcanic products.

Abstract

Copahue volcano is an active stratovolcano located in the Southwest of the Agrio-Caviahue depression in the Province of Neuquén at the border between Argentina and Chile, and the most recent eruptive activity involved five cycles between 1992 and 2016. The Moderate Resolution Imaging Spectroradiometer (MODIS) is an instrument onboard the satellites Terra and Aqua that provides, on average, four images per day and has channels centered at 4 μm and 12 μm that are useful for detecting volcano temperature anomalies. In this work, we evaluated the normalized temperature index (NTI) as implemented by MODVOLC and the radiance recorded by MODIS at 4 μm as measures of precursory volcanic activity between October 2011 and January 2013, with a special focus on the December 2012 eruption. We evaluated the thermal anomalies identified by MODVOLC based on a global threshold of −0.8 and observed that they were namely syn- or post-eruptive and mostly correlated with the presence of incandescence material related to Strombolian activity. The global nighttime threshold was adapted to the local context of Copahue and we observed that a threshold of −0.875 is a more suitable value for the examined period. An analysis of time series of the NTI and of the radiance at 4 μm based on 116 MODIS images, resulted in that the adapted threshold allowed the detection of anomalies undetected by the global value. The anomalies were in either case unrelated to precursory activity but correlated with surface expressions of activity such as the emission of volcanic ash, volcano degassing, and the ejection of ballistic and incandescent material. The knowledge on the chronology of activity prior to and during the eruption could be improved, as regards the thermal behavior of the volcano in relation also to geochemical, deformation, and seismic signals. The replication of this kind of analysis in other periods of activity at Copahue and other volcanoes, the evaluation of other algorithms and the incorporation of data acquired by other satellite instruments are necessary to improve our knowledge on the temperature behavior of volcanoes before and during periods of unrest. This would ultimately contribute towards the improvement of volcano monitoring capabilities in volcanic areas.

Introduction

Volcanic eruptions involve the transfer of matter and energy from the interior to the surface of the Earth. This process can be measured from space by infrared sensors on board satellites. Volcanoes are sometimes located in areas of difficult access and monitoring active or potentially active volcanoes require the use of different techniques to understand these complex systems and mitigate the hazards and risks they pose to communities. Remote sensing complements ground-based volcano monitoring techniques, because of the low cost and adaptability in relation to the ability it provides to cover large areas, while minimizing the exposure to volcanic hazards (Dean et al., 1998; Oppenheimer and Rothery 1991). Satellite sensors facilitate the effective management of the different phases of volcanic crises, from unrest to post-eruption activity (Pyle et al., 2013). In fact, space-borne data can be used for the early evaluation of thermal alerts that may act as precursory signals of activity at the local (Murphy et al., 2013; Coppola et al., 2015b, 2016; Laiolo et al., 2017; Candela-Becerra et al., 2020), regional (Dehn et al., 2000; Rothery et al., 2005; Jay et al., 2013; Layana et al., 2020) and global scales (Wright et al., 2002). Temperature anomalies measured from space enable the determination of the phases of activity that may indicate possible eruptions when analyzed together with seismic, geochemical and deformation data (Henderson et al., 2019; Reath et al., 2020).

Radiometers on board satellites have been used since the 80s to address volcano thermal processes and anomalies including the detection of hot spots, the analysis of surface thermal structures, the estimation of heat and mass flux rates, the description and interpretation of eruption chronologies and the analysis of time series (Harris, 2013). Thermal anomalies are portions of the surface that differ in the radiance emitted, when compared to the emission of the background and depending on the magnitude of the radiance, they can be classified as high or low temperature anomalies (Flynn et al., 2000; Murphy et al., 2013; Harris, 2013). The measurement of low temperatures near ambient require instruments with channels centered at thermal infrared (TIR) wavelengths, i.e. 5.0–15 $\mathrm{\mu m}$, whereas high temperature anomalies must be addressed with instrument bands centered at the middle (MIR) and shortwave infrared (SWIR) wavelengths, i.e. 3.0–5.0 $\mathrm{\mu m}$ and 1.1–3.0 $\mathrm{\mu m}$, respectively. These sensors may be of moderate to high spatial resolution with a pixel size of 60–100 m on the ground and low revisit time, such as the thermal infrared sensor onboard the satellite Landsat 8 or the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) onboard the satellite Terra. Alternatively, infrared radiometers may be of moderate to low spatial resolution with a pixel size of 0.5–1.0 km at nadir and high revisit time, such as the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard the Terra and Aqua satellites, or the Visible Infrared Imaging Radiometer Suite (VIIRS) on board the Suomi National Polar-orbiting Partnership (Suomi NPP) and NOAA-20 weather satellites. Finally, instruments on board geo-stationary orbit satellites such as GOES-16 ABI, which has a field of view that covers the American continent, or the Spinning Enhanced Visible and Infrared Imager (SEVIRI) on board Meteosat Second Generation (MSG), which covers Europe and Africa, are able to observe the Earth in multiple spectral channels with a frequency of 15 min or less and at a spatial resolution of 2–3 km. In this study, we used MODIS to investigate the thermal anomalies at Copahue volcano focused on the December 2012 eruption. Identical versions of MODIS are on board the satellites Terra and Aqua with their first acquisition in February 2000, both with a sun -synchronous orbit; the cross-track swath width is 2330 km and images are acquired twice a day within a 24 h period, during daytime and nighttime, respectively. Data is acquired across 36 spectral channels covering the visible and near infrared (VNIR), the short-wave infrared (SWIR), the middle-wave infrared (MIR) and the thermal infrared (TIR) portions of the electromagnetic spectrum. Their spatial resolution encompasses resolutions of 250 m (bands 1 and 2), 500 m (bands 3 to 7) and 1 km (bands 8 to 36), with accuracy of geolocation of less than 1 km (Wright et al., 2002).

The systems for the automatic identification of high temperature anomalies comprise the implementation of algorithms to enable an objective way to process large volumes of data generated by satellites with a high revisit period and channels centered in the middle and thermal infrared. This kind of systems facilitate the analysis of changes in thermal parameters related to volcanic activity (Harris, 2013; Ramsey and Harris, 2013), the identification of volcanic unrest (Laiolo et al., 2017; Furtney et al., 2018), the forecast of eruptive trends (Coppola et al., 2015b, 2016) and in turn contribute towards the improvement of volcano monitoring capabilities at active and remote volcanoes (Coppola et al., 2020; Reath et al., 2020). The identification of high temperature anomalies can be done in two ways, either by visual inspection of the radiance at 4 μm (Rad. 4 μm) or by means of an automatic or semi-automatic procedure based on a detection algorithm that exploits the spectral, spatial and/or temporal characteristics of an image (Steffke and Harris, 2011; Harris, 2013). Hotspot detection algorithms may rely on the application of a fixed threshold to the spectral Rad. 4 μm or to a spectral index that exploits the differential response of anomalous and background pixels at 4 and 12 μm. MODVOLC implements the Normalized Thermal Index (NTI), which is a ratio of the radiances measured by MODIS at both wavelengths (Wright et al., 2002, 2004) and applies a day and nighttime threshold of −0.6 and −0.8, respectively, in order to identify volcanic temperature anomalies at a global scale. Anomalies are published in an open access website (http://modis.higp.hawaii.edu). Some authors, however, have modified these thresholds to a local context to facilitate the identification of features and processes that would go otherwise undetected (e.g. Pergola et al., 2004; Kervyn et al., 2008; Murphy et al., 2011). Contextual algorithms consider the spatial statistics of the radiance or temperature of the background. This is typically defined by the eight pixels surrounding a central target one (e.g. VAST was the first one designed for the detection of anomalies using data acquired by the Advanced Very High Resolution Radiometer (AVHRR) (e.g. Harris et al., 1995; Harris et al., 1997; Harris, 2013). Other systems may characterize the background by using statistical measures (i.e. mean and standard deviation) of time series of satellite data acquisitions such as the Robust Satellite Technique (RST) (Tramutoli 1998; Pergola et al., 2004; Steffke and Harris, 2011; Harris, 2013). Finally, hybrid approaches may combine more than one of the three dimensions of the data mentioned above. Koeppen et al. (2011) improved the capabilities of MODVOLC by incorporating a time series component into the algorithm by using the Robust Satellite Technique (Tramutoli 1998; Pergola et al., 2004). MIROVA (Middle Infrared Observation of Volcanic Activity) combines the NTI (Wright et al., 2002), the Enhanced Thermal Index (ETI) and considers the spatial context (Coppola et al., 2015a) to identify anomalies and estimate the radiative power at 216 active volcanoes using MODIS data (Coppola et al., 2020). Lastly, REALVOC (Kaneko et al., 2010) uses MODIS channels centered at 4 and 11 μm (i.e. bands 21 and 31, respectively), and the Multi-Functional Transport Satellite (MTSAT) for monitoring anomalies and tracking eruptive events in near real time at 147 volcanoes in East Asia and some volcanoes in North of South America (Castaño et al., 2020).

The aim of this paper is to evaluate the NTI and the Rad. 4 μm as possible indicators of volcanic activity and assessing the use of an adapted threshold to identify temperature anomalies related to the precursory thermal activity. We first evaluated the anomalies identified by MODVOLC and published online between 2000 and 2016 and then performed a comprehensive analysis of the NTI and the Rad. 4 μm between October 2011 and January 2013, i.e. prior to and during the December 2012 eruption. We selected the eruptive cycle that started in December 2012 until the end of January 2013 due to its intensity and duration. We adapted MODVOLC's threshold to the local context of Copahue by using image statistics of MODIS nighttime data. We evaluated both thresholds through the analysis of time series of the NTI and Rad. 4 μm at two frequencies of observation: (i) with one image per week between October 2011 and November 2012, i.e. before the eruption, and (ii) with images acquired every day in December 2012 and January 2013, i.e. just before, during and just after the eruption. We investigated the behavior of both parameters at the pixel that includes the crater lake and an area of 10 × 10 km around it and attempted to correlate the presence of anomalies according to the global and adapted thresholds with expressions of volcanic activity. We were therefore able to test the hypothesis of the NTI and the Rad. 4 μm as indicators of precursory activity; we also reconstructed the chronology of the activity of Copahue volcano prior to and during the December 2012 eruption and discussed the genesis of the activity in relation to geo-chemical, deformation and seismic signals.