Unmanned aerial vehicles (UAVs) as a tool for hazard assessment: The 2021 eruption of Cumbre Vieja volcano, La Palma Island (Spain)

Unmanned aerial vehicles (UAVs) as a tool for hazard assessment: The 2021 eruption of Cumbre Vieja volcano, La Palma Island (Spain) A. Román a, ⁎,A.Tovar-Sánchez a , D. Roque-Atienza a ,I.E.Huertas a ,I.Caballero a ,E.Fraile-Nuez b , G. Navarro a a Department of Ecology and Coastal Management, Institute of Marine Sciences of Andalusia (ICMAN), Spanish National Research Council (CSIC), 11510 Puerto Real, Spain b Canary Islands Oceanographic Centre, Spanish Institute of Oceanography (IEO), Spanish National Research Council (CSIC), 38180 Santa Cruz de Tenerife, Spain HIGHLIGHTS •UAVs provide crucial information for hazard assessment during a volcanic event. •Feasible tool for accurately elaborate DEMs, 3D models and topographic maps •Centimeter-scale approximation of the area devastated by the lava flow •UAV data allowed the performance of lava flow flooding forecasting models. •Chemical analysis of surface water samples obtained under the influence of the lava delta GRAPHICAL ABSTRACT ABSTRACTARTICLE INFO Editor: Damià Barceló Keywords: Unmanned aerial vehicle (UAV) Volcanic eruptions Hazard assessment La Palma eruption Lava forecasting Monitoring for assessment of natural disasters, such as volcanic eruptions, presents a methodological challenge for the scientific community. Here, we present Unmanned Aerial Vehicles (UAVs) as a feasible, precise, rapid and safe tool for real time monitoring of the impacts of a volcanic event during the Cumbre Vieja eruption on La Palma Island, Spain (2021). UAV surveys with optical RGB (Red-Green-Blue), thermal and multispectral sensors, and a water sampling device, were carried out in different areas affected by the lava flow, including the upper volcanic edifice and the lava delta formed on the coastal fringe of the island. Our results have provided useful information for the monitoring of the advance of the lava flow and its environmental consequences during the volcanic emergency. Our data shows how La Palma island's growth, with the formation of a new lava delta of 28 ha and a total volume of lava injected into the sea of 5,138,852 m 3 . Moreover, our Digital Elevation Model (DEM) simulated, with a 70 % accuracy, the probabilistic simulation of the possible path followed by the lava flow in the vicinity of the fissure from which the magma emanates. In addition, significant changes of seawater physical-chemical parameters were registered in coastal surface waters by the in situ seawater samples collected with the automatic water sampling device of our UAV. The first meters of the water column, due to the instant evaporation of the seawater in contact with the hot lava, produce an increase of temperature and salinity of up to 4–5 °C and up to 5 units, respectively. 1. Introduction A natural disaster constitutes a dynamic scenario, which is complicated to manage and has negative impacts on the natural environment, economy and society (Colson et al., 2018;Dou et al., 2014;Giordan et al., 2018). Consequently, a large number of people are affected by natural disasters in both developing and developed economies (Giordan et al., 2018;Krapivin et al., 2012;Kucharczyk and Hugenholtz, 2021;Sukov et al., 2008). For this reason, suitable methodologies are essential in order to evaluate the impact of these phenomena and in particular to monitor their evolution in real time in order to help decision makers minimize and/or mitigate the damages Science of the Total Environment 843 (2022) 157092 ⁎Corresponding author. E-mail address: [email protected] (A. Román). http://dx.doi.org/10.1016/j.scitotenv.2022.157092 Received 10 May 2022; Received in revised form 8 June 2022; Accepted 27 June 2022 Availableonline30June2022 0048-9697/© 2022 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Contents lists available at ScienceDirect Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv during emergencies (Cracknell and Varotsos, 2011;Giordan et al., 2018). The use of satellite imagery for the monitoring of natural disasters has become a common practice, since satellites operate continuously covering wide areas with medium-high spatial resolution (Ganci et al., 2020; Solikhin et al., 2012). The application of remote sensing satellite imagery for natural disasters began around 2000 by the International Charter Space and Major Disasters and the United Nations (UN) (Bessis et al., 2004;Kaku, 2019). Nowadays, global emergency agencies using satellitebased monitoring support the management of natural disasters by providing geospatial information derived from optical instruments, thermal sensors, altimetry systems, synthetic aperture radar (SAR), Laser Imaging Detection and Ranging (LiDAR), radiometers and spectrometers (Borgeaud et al., 2015;Colson et al., 2018). In particular, these agencies include the United States Geological Surveys (USGS) Emergency Response (Fink et al., 1990;Mueller et al., 2015), the Asia-Pacific Regional Space Agency Forum (APRSAF) Sentinel Asia (Adriano et al., 2019;Kaku, 2019), or the Copernicus Emergency Management Service (CEMS) (Copernicus Emergency Management Service, n.d.), whose space segment is made up of five Sentinel satellites that allow the monitoring of floods (DeVries et al., 2020;Konapala et al., 2021;Yuan et al., 2021), earthquakes (Jelének and Kopacková-Strnadová, 2021;Li et al., 2021), harmful algal blooms (Caballero et al., 2020;Ogashawara, 2019;Rodríguez-Benito et al., 2020), wildfires (Ban et al., 2020;Colson et al., 2018;Llorens et al., 2021) and volcanic eruptions (Bell et al., 2021;Coca et al., 2014;Gray et al., 2019;Plank et al., 2020), among others. Volcanic eruptions are hazardous and particularly challenging for ground-data collection or direct observations (Walter et al., 2020), and require remote sensing tools for their monitoring and risk assessment. However, applicability of satellite remote sensing in the study of volcanic events remains limited because of the reliance on optimal meteorological conditions and the consequent unavailability of observations for specific dates and locations (Andaru et al., 2021;Bonali et al., 2019), with the exception of those equipped with thermal and radar sensors (e.g. SAR), which can work at night, with cloud cover, etc. In addition, despite the advances in satellite remote sensing technology, it is still difficult to monitor these dynamic phenomena due to their coarse spatial image resolution and their long re-acquisition intervals (Turner et al., 2017). To overcome these limitations, the use of different sensors mounted on Unmanned Aerial Vehicles (UAVs) for the monitoring of volcanic eruptions is gaining prominence as a complement to satellite remote sensing (Andaru et al., 2021; Pering et al., 2020;Zorn et al., 2020). There are few studies that have employed UAVs in the assessment of volcanic eruptions assessment to: evaluate their geomorphological description and elaborate digital terrain models (Darmawan et al., 2018;Favalli et al., 2018;Walter et al., 2020); measure the volcanic gas composition of the ash column (Kazahaya et al., 2019;Liu et al., 2019;Mori et al., 2016); analyze the lava flow characteristics and develop models that allow the prediction of their behavior (Dietterich et al., 2017;Turner et al., 2017); or sample water at crater lakes (Terada et al., 2018). UAVs offer an alternative to overcome the main limitations of other remote sensing techniques such as satellites during volcanic crises, since: i) they provide topographic data with centimeter spatial resolution; ii) they allow better control of temporal resolution, increasing the flexibility for image/video acquisition or water/gas sampling; iii) their operation is not limited by the presence of volcanic ash within eruption columns; iv) they reduce human risk during data collection in the field, particularly in difficult, restricted or dangerous areas; v) multiple parameters can be measured with different devices or sensors (thermal, hyperspectral, multispectral, RGB, LiDAR, gas detector, water sampling device); and vi) they are able to obtain results in real time facilitating the topographic modeling and lava flow dynamics for hazard evaluation (Giordan et al., 2018; Gomez and Purdie, 2016;Turner et al., 2017). Here, we present the La Palma volcanic eruption (Canary Islands, Spain) as our case study (Fig. 1). The new eruption on the Cumbre Vieja ridge, started on September 19th 2021 from multiple vents aligned in a single linear fissure segment in the Earth's crust, emitting (inland and towards the ocean) large amounts of lava and volcanic material during the eruption period. The eruption was officially declared finalized on December 25th 2021, after three months of spewing ash and hot molten rock. Volcanic eruptions directly affect people living on or near the volcano, who are forced to be evacuated due to: i) ash fallout, that may cause building collapse as a result of rapid accumulation and high density of volcanic ash particles; ii) toxic gases (SO 2 ,H 2 S) which may pose respiratory health risks; or iii) direct effect of lava flow destroying numerous hectares of cultivated land and buildings (Malawani et al., 2021). In addition, lava flows from the eruption can lead to geomorphological transformations in the volcano edifice due to erosion and sedimentation processes, and they have a significant impact on water bodies (including lakes, reservoirs or the ocean) since they affect water quality, generate harmful gases on contact with water, or even create small tsunamis (Malawani et al., 2021;Stewart et al., 2006). In order to monitor the impacts and environmental processes occurring during the eruption, we used two UAVs equipped with different sensors, as well as a water sampling device (see Methods section), to provide real time data of the evolution of the volcanic activity. We elaborated Digital Elevation Models (DEMs), orthophotos and 3D models from structure from motion (SfM) photogrammetry during three eruption days, allowing us to analyze the morphology of sub-meter-scale volcanic features over kilometer-scale areas; to detect temperature anomalies that can be indicative of magmatic activity; to explore the lava flow path and characteristics in order to contribute to hazard assessment; and to determine the impact of the magma on the ocean. The survey also aims to provide precise data for emergency teams and authorities for the management of the effects caused by the eruption in real time, highlighting the potential of UAV technology to rapidly obtain high-spatial-resolution data during volcanic crises. 2. Methods 2.1. Study area During the fieldwork on October 1st-3rd 2021, three different areas of the Cumbre Vieja volcanic eruption on La Palma Island (28°40′0.01″N, 17°52′0.01″W) were surveyed in order to acquire visible, multispectral and thermal imagery using different sensors on board UAVs, as well as surface water samples (Fig. 1). The first two monitored areas were close to the fissures from which the magma emanated to the surface (vent region), corresponding to the upper region of the lava flows (Fig. 1d, e). The third monitored area corresponds to the lava delta formed on the coast of Tazacorte (Fig. 1c). The volcano erupted on September 19th 2021 with a fissuraltype eruption, and the lava reached the sea ten days later. Approximately 700 buildings were destroyed by the lava flow according to the data provided by the Council of La Palma (Council of La Palma, n.d.) and the Copernicus Emergency Management Service (CEMS) (Copernicus Emergency Management Service, n.d.) during the flight date. Nevertheless, and due to the continuous advance of the lava, the total number of destroyed buildings continued rising every day, reaching a total of 2988 buildings covered by the flow at the end of the eruption (Copernicus Emergency Management Service, n.d.). After 85 days of eruption, December 25th 2021 was considered the end of the volcanic event, the longest eruption recorded in the history of the island (IGN, 2021). 2.2. Sentinel-2 imagery The Sentinel-2 twin polar-orbiting satellites developed by the European Commission (EC) and the European Space Agency (ESA) were also used in this study. The multispectral instruments (MSI) on-board both satellites are now operational with a global 5-day revisit frequency at the equator. The MSI samples 13 spectral bands: four bands at 10 m, six bands at 20 m and three bands at 60 m spatial resolution. Detailed information about the radiometric and spectral characteristics of the visible and near-infrared (NIR) bands are specified in the User Handbook (ESA, 2015, 2017). The Sentinel-2 Level-2A surface reflectance products (corrected from the A. Román et al. Science of the Total Environment 843 (2022) 157092 2 atmospheric effects) at 10 m spatial resolution used in this study corresponded to the September 10th 2021 and the October 15th 2021 scenes (tile 28RBS). 2.3. UAV platforms and sensors In this study, two UAVs were used: 1) A hexacopter with three-bladed propellers (Condor, Dronetools ©), which has a DJI6010 electric motor powered by four Li-ion batteries (7000 mA each). The empty weight of the equipment with the four batteries is 11.8 kg (maximum takeoff weight (MTOW) is 14.9 kg), with maximum flight autonomy of up to 60 min (without payload). This UAV was simultaneously equipped with a MicaSense RedEdge-MX dual multispectral camera and the radiometric thermal camera FLIR Vue Pro 19 mm. The multispectral camera has 10 different spectral bands able to acquire data in the following wavelengths: coastal blue 444 nm, blue 475 nm, green 531 and 560 nm, red 650 and 668 nm, red edge 705, 717 and 740 nm, and near infrared (NIR) 842 nm. In addition, the camera allows a broad surface coverage since it has a horizontal field of view of 47.2°, and a spatial resolution of 8 cm/pixel from 120 m high. It also has a Downwelling Light Sensor (DLS) with built-in GPS providing more accurate and reliable measurements of irradiance and solar angle, and the included calibration panel (RP041924106-0B) for improved radiometric calibration. The thermal camera captures non-contact temperature measurements with calibrated temperature data embedded in every pixel, its spectral band is 7.5–13.5 μm and it has a sensor resolution of 336 × 256. In addition, this UAV was equipped with a water sampling device (Sparaventi et al., 2022) with an acid cleaned plastic weighted container (2 L) and a castable instrument hanging from the UAV with a 5-meter rope providing instantaneous profiles of temperature and salinity (CastAway R -CTD) and collected surface water. 2) A quadcopter DJI Mavic 2 Pro with an RGB Hasselbald L1D-20c camera with 1”CMOS and 20 MP. Weighing 907 g, it has a maximum flight autonomy of 31 min, which is usually the limiting factor when flying over large areas. 2.4. UAV operations The flights were preprogrammed using the software DJI Ground Station Pro (GSP), after verifying that weather conditions were favorable for the flight and considering geographical factors, such as the elevation of the terrain. Constant parameters were previously established, including flight height, flight speed, flight time, ground sampling distance (GSD), the trajectory of the UAV and the overlaps (always 80 % frontlap and 70 % sidelap). The multispectral sensor was calibrated, before and after the flight, using a referenced reflectance panel. The Spanish civil aviation regulations for an emergency situation, for which the Spanish Agency for Aviation Safety (AESA) is responsible, were followed during all operations deployed on the ground. All UAV operations of this campaign were carried out between October 1st and 3rd 2021. The software Pix4D Mapper (Pix4D © SA, Lausanne, Switzerland, v.4.6.4, https://www.pix4d.com/) was used for the SfM photogrammetry. Once the images were imported into the software, different processing steps were performed, including the conversion of digital numbers into relative temperature (°C) for thermal imagery, the generation of the point cloud, the generation of the textured 3D mesh, where obtained 3D points are interpolated to form a triangulated irregular network in order to obtain Digital Elevation Model (DEM). This DEM is then used to project every image pixel and to calculate the georeferenced orthomosaic. The photogrammetric products generated were: i) a 15.7 cm/pixel orthomosaic, a 15.7 cm DEM and a 3D model composed of 60 photos taken at the lava flows adjacent to the vent region at 400 m altitude by an optical RGB sensor (DJI Mavic 2 Pro); ii) a 8.8 cm/pixel orthomosaic and a 9 cm DEM resulting from 176 photos taken at the upper part of the lava flows by an optical RGB Fig. 1. a) Sentinel-2 level 2A false color (urban) composite (bands 12, 11, 4) from September 30th 2021, with b) detailed representation of the three volcano areas covered in this study. Flight plans over c) the lava delta formed over the sea, d) the upper region of lava flows and e) the vent region and the adjacent lava flows in the flank of the cone. Pictures of f) the main eruptive vent and g) the lava delta taken October 1st 2021. The QGIS software (v.3.16.14, https://qgis.org/downloads/) was used to create the figures. A. Román et al. Science of the Total Environment 843 (2022) 157092 3 A. Román et al. Science of the Total Environment 843 (2022) 157092 4 sensor (DJI Mavic 2 Pro) in a 300 m altitude flight; iii) a 3D model and a 253 photo orthomosaic of 11.8 cm/pixel resolution generated from three optical RGB flights (DJI Mavic 2 Pro) repeated with different gimbal angles (60° and 90°) at a constant height of 300 m over the lava delta; and iv) an orthomosaic of 25.2 cm/pixel derived from 357 thermal captures and an orthomosaic of 20.7 cm/pixel for each of the 10 multispectral bands from a total of 3430 photos taken at the lava delta at 200 m altitude. Finally, the QGIS (QGIS Development Team, Geographic Information System, Open Source Geospatial Foundation Project, v.3.16.14, https:// qgis.org) and SAGA GIS (Conrad et al., 2015, v.7.9.0, https://saga-gis. sourceforge.io/en/index.html) programmes were used to generate topographic models and for their subsequent analysis. Fig. 1 was generated based on the georeferenced information derived from the UAV orthomosaics and the flight plan point clouds obtained from Pix4D processing. The information provided by the DEMs generated from the UAV flights, allowed the elaboration of topographic maps in QGIS using the “slope”raster analysis tool (Fig. 2), and the raster calculator to measure surface areas and volumes (Fig. 4). Quantum-Lava Hazard Assessment (Q-LavHA) was theQGISpluginusedforthelavaflow invasion probability simulation (Fig. 3). It combines existing probabilistic and deterministic models (Felpeto et al., 2001;Harris and Rowland, 2001) and proposes some improvements to calculate the probability of lava flow spatial propagation and terminal length. The established parameters for the simulation execution were obtained from the DEMs elaborated with the UAV flights. Specifically, the DEMs themselves, the geographic locations (in UTM) of the volcanic vents, and topographic data such as the value of the elevation in the pixel where the lava flow runs were used. Image classification (Fig. 5) was performed with the supervised classification technique “support vector machine”(SVM), a machine learning algorithm that can successfully handle data with unknown statistical distributions and with small training sets created from spectral data and validated with in-situ information (Vapnick, 1995). This approach applies kernel functions that map the training data into a larger-dimensional space in which classes can be linearly separated by a hyperplane. In this case study, the radial base function was used as the transformation nucleus, since it offers the best results as reported by Miranda et al. (2020). Four different land cover classes were created based on the spectral information obtained from vegetation, water, lava flow and any element other than the three previous classes, respectively. The QGIS raster calculator and the SCP (Semiautomatic Classification Plugin) were used for UAV band composites and index generation (NDVI) in Fig. 5. 2.5. Chemical analysis Total alkalinity (TA) was determined by titration of seawater using a potentiometric system (Mintrop et al., 2002) with a Metrohm 794 Titroprocessor. Water samples were taken directly from the water sampling device and stored in 500 mL borosilicate bottles and poisoned with 100 μL of a saturated aqueous solution of mercuric chloride for later shore-based analysis. The accuracy of the TA determinations was assessed by measurements of Certified Reference Material (CRM batch # 186, supplied by Prof. Andrew Dickson, Scripps Institution of Oceanography, La Jolla, CA, USA), being ±0,06 μmol/kg. Seawater pH was measured using a Metrohm 780 pH meter equipped with a combined glass electrode (Metrohm model 6.0258.010) that included a temperature probe, which was calibrated following the protocol described by Del Valls and Dickson (1998),withanaccuracy of ±0.004 pH units. The pH values were then obtained in SWS scale and subsequently transformed to total scale and referred to the in-situ temperature (pH T ) using the CO2sys.xls programme (Lewis et al., 1998). The partial pressure of CO 2 ,pCO 2 , was also computed afterwards using the Fig. 3. Lava flow simulation in the vent region and adjacent lava flows in the flank of the cone with the Q-LavHA model from the DEM produced with the UAV flight on October 3rd 2021. The model prediction is shown in red scale, while the real lava flow is shown in yellow. The QGIS software (v.3.16.14, https://qgis.org/downloads/) was used to create the figure. Fig. 2. Topographic models generated with the SfM workflow followed for the DJI Mavic 2 Pro flight performed on October 1st (at constant 300 m altitude) and 3rd (at constant 400 m altitude) 2021. a) Optical RGB orthomosaic of the vent region and the proximal lava flows in the flank of the cone; b) slope map containing contour lines between 600 and 900 m altitude every 25 m, and the slope expressed in °; c) optical RGB orthomosaic of the upper region of lava flows; d) 225 m topographic profile A– B showing the effect of A’a-type lava flow on the terrain; e) 80 m topographic profile C–D showing the effect of Pahoehoe-type lava flow on the terrain; f) 3D model of the volcanic structure. The QGIS software (v.3.16.14, https://qgis.org/downloads/) was used to create the figures. A. Román et al. Science of the Total Environment 843 (2022) 157092 5 pair pH T , TA values, and proper dissociation constants (Dickson and Millero, 1987;Mehrbach et al., 1973). Seawater samples for the determination of DOC and TDN were collected in 0.25 L acid–cleaned glass bottles and immediately filtered through precombusted (450 °C, 4 h) Whatman GF/F filters with an acid-cleaned all glass filtration system. Aliquots of 20 mL were collected for DOC analysis in precombusted (450 °C, 12 h) 24 mL glass vials. After acidification with H3PO4 (85 %) to pH <2, they were sealed with Teflon-lined caps and stored in the dark at 4 °C until analyzed in the shore-based laboratory within 1 week of collection. DOC content in samples was measured with a commercial Shimadzu TOCL/CPH/ CPN organic carbon analyzer working under the principle of hightemperature catalytic oxidation (Salgado and Miller, 1998). The precision of the analyzer was ±0.5 μmol L −1 . Accuracy of measurements was checked with CRM provided by D. A. Hansell (University of Miami, USA), resulting in ±1.3 and −1.8 for DOC and TDN, respectively. 3. Results DEMs obtained with UAV photogrammetry are essential to assess the lava flow morphology and to obtain several parameters used for flow modeling in the vicinity of the volcanic vent. Lava flows are fundamentally gravity currents that are modulated by internal factors such as viscosity or inflation rates, and by external factors such as slope, terrain roughness or other topographical elements (DeGraffenried et al., 2021;Gomez and Purdie, 2016;Turner et al., 2017). Slope plays a critical role in controlling lava flow behavior, as steeper slope drives advance rates, while topographical features can plug or slow flows (Gomez and Purdie, 2016). Pahoehoetype lava flows generally move slower than A’a-type flows due to their high density and thickness, although they can cover larger areas (Turner et al., 2017). In the La Palma eruption, the lava flow from the volcano was mainly of A’a-type and Pahoehoe-type (IGN, 2021), as visually observed in the photographs taken in-situ and in video footage that shows how the lava flows through the upper flank of the cone (see Supplementary Movie 1 file). Fig. 2 shows the topographic products resulting from the DEM analysis, which allowed the elaboration of optical RGB orthomosaics (Fig. 2a–c), a slope map with contour lines (Fig. 2b), topographic profiles (Fig. 2d–e), and a 3D model (Fig. 2f) that would allow visual confirmation of the values indicated on the slope map. These products allowed us to obtain valuable parameters for lava flow modeling, such as the slope or terrain elevation, and information on the pathway followed by the lava flow, with a visible decrease of the ground cone from 900 m to 600 m altitude. Similarly, the topographic profile A–B(Fig. 2d) indicates an elevation of approximately 2 m corresponding to the A'a-type lava flow, while the topographic profile C–D(Fig. 2e) shows the elevation of the terrain associated with the accumulation of high density rubbly Pahoehoe-type lava materials, which in some areas reached a height of 12 m. The slope map reflects a more abrupt and irregularly lobulated morphology with slopes between 6° and 15°. The highest slopes were found around the vent region, with approximate values of 25°. A simple probabilistic flow model (Quantum-Lava Hazard Assessment, Q-LavHA) was generated in order to predict possible lava flow paths and to provide flow hazard assessments (see Methods section). In this case, the DEM obtained during the volcanic event of the most proximal area to the vents through which the lava flow runs, was used. Therefore, it is possible to compare the prediction with real data of the trajectory followed by the lava flow, and demonstrate how these data help predict the most likely lava pathway, thus facilitating prevention and protection activities for emergency services. Fig. 3 shows the areas with the highest probability of flooding due to lava flow, based on the simulation carried out by the QLavHA probabilistic model for QGIS. There is a correspondence of over 70 % between the real trajectory followed by the lava flow and the model prediction. An A’a-type lava delta was formed on the west cliff of the island, near the municipality of Tazacorte. A lava delta is a fan-shaped accumulation of lava flow materials that occurs when the magma that comes into contact with seawater rapidly solidifies, growing in size as volcanic material is accumulated (Bosman et al., 2014;Di Traglia et al., 2018;Zhao et al., 2020). In addition, seawater boils when interacting with lava, generating a toxic column of water vapor, hydrochloric acid and pyroclastics that can reach heights and distances of tens of meters (Hildenbrand et al., 2012;Poland and Orr, 2014). The importance of the study of lava deltas lies not only in the topographic modification of the total surface of the volcanic island (which is increased), but also in its instability since its composition and morphology favor catastrophic collapse that can generate coastal explosions of hurl rocks and fine ash, or even small tsunamis (Bosman et al., 2014;Poland and Orr, 2014). Fig. 4a–b shows the change in the coastline that took place after the formation of the A’a-type lava delta, with a true color composite of Sentinel-2 Level 2A pre-eruption (September 10th 2021) and another false color composite post-eruption (October 15th 2021). The 3D model (Fig. 4c) and optical RGB orthomosaic (Fig. 4d) of the lava delta's main body, allowed the determination of 28 ha (Fig. 4e) over the ocean on October 2nd 2021, and was characterized by an overall lobate morphology with slope gradients between 20°–40°. However, the delta continued growing until it reached approximately 33 ha on October 15th 2021, remaining stable according to the data provided by the CEMS (Copernicus Emergency Management Service, n.d.). It is also evident that there was a higher accumulation of lava at the falling surface of the cliff with abrupt lobes being observed on the flanks, leaving a short steep slope inside the main delta body that rises a few meters above sea level. The DEM analysis allowed us to obtain an approximate total volume of 5,138,852 m 3 in the lava delta, this accumulation being larger in the 4 ha closest to the magma fall at the edge of the cliff, with a volume of 297,220 m 3 . In addition, a photogrammetric flight was carried out in this area with an UAV equipped simultaneously with thermal and multispectral sensors (see Methods section). The thermal map (Fig. 5a) shows the relative temperature changes of the lava flow, colder in the area of the lava delta where it solidified after contact with seawater, and much hotter (the thermal sensor saturates above 150 °C) on the cliff. The false color composite (near infrared (NIR) 840 nm, green 560 nm and red 668 nm bands) and the Normalized Different Vegetation Index (NDVI) (Fig. 5b–c, respectively) show the presence of banana plantations in the areaaffected by the volcanic activity, which is the most important economic activity on the island. A supervised classification (Fig. 5d) in four ground cover classes was carried out using the “Support Vector Machine”(SVM) method (Vapnick, 1995, see the Methods section), with an overall classification accuracy of 92.07 % and a Cohen's Kappa Index of 0.88 (see uncertainties and error matrix in Supplementary Files Section, Table A). The thematic map shows that 17 ha of banana plantations of the 35 ha identified by Sentinel-2 data (Fig. 4a), were razed to the ground by the lava in the study area. The volcanic eruption has not only severely impacted buildings and infrastructure of the island, but also the marine environment, with physical-chemical and biological alterations of coastal seawater, which can affect fisheries, that contribute significantly to the island's economy. It is well known that volcanic events provoke explosions generated by thermohydraulic reactions, involving high rates of heat transfer between lava and seawater, and chemical reactions, and the release of huge amounts of metals and gases, mainly Fe, Cu, Cd, Hg and chlorine (with small amounts of sulfur species) that alter the chemistry of seawater (James et al., 2020;Mason et al., 2021). The water sampling device mounted on board the UAV (Tovar-Sánchez et al., 2021,seeMethods section) allowed the collection of surface water samples in the areas adjacent to the lava delta, located inside a 500 m security navigation perimeter established by the authorities and, therefore, not accessible by vessels. Fig. 6 shows temperature and salinity profiles that were obtained at six stations located in the vicinity of the lava delta. The results obtained for the first seven meters of the water column reveal a temperature rise (up to 4–5 °C) and a salinity increase (up to five units) in the first meter due to the instant effect of the hot lava on the colder ocean and the instant evaporation of the seawater, respectively. These important thermohaline anomalies A. Román et al. Science of the Total Environment 843 (2022) 157092 6 Fig. 4. a) Sentinel-2 level 2A true color composite (bands 4, 3, 2) from September 10th 2021 (pre-eruption);b) Sentinel-2 level 2A false color (urban) composite (bands 12, 11, 4) from October 15th 2021 (post-eruption). Topographic models generated with the SfM workflow followed for the DJI Mavic 2 Pro flight performed on October 2nd 2021 at constant 300 m altitude over the lava delta: c) 3D model; d) optical RGB orthomosaic of the lava delta; e) area of the lava delta in hectares measured over the optical RGB orthomosaic. The QGIS software (v.3.16.14, https://qgis.org/downloads/) was used to create the figures. A. Román et al. Science of the Total Environment 843 (2022) 157092 7 highlighted the volcanic influence on ocean chemistry since these large variations in temperature and salinity are atypical in the ocean. Table 1 summarizes the values of carbon system parameters and total dissolved nitrogen in surface waters closed to the lava delta, where Blank represents values found at a reference station located 600 m away from the lava delta and, therefore, unaffected by the lava cascade. In this latter case, pH T and pCO 2 values that are referred to as in-situ conditions throughout the text, along with total alcalinity (TA) values, were similar to those typically observed in the Canary basin (González-Dávila et al., 2006). In contrast, surface seawater chemistry around the lava delta was highly altered, with pH T and TA values decreasing in all the sampled sites with respect to those measured in the remote station. In particular, pH T ranged from 7.8896 in D4 to 7.2473 in D6, with this latter station and D2 exhibiting pH T values that were 0.7 and 0.57 units lower than the pH T in the unaffected area (7.9516). This was also the trend encountered for the TA, as this variable diminished in the surroundings of the lava delta with respect to the TA measured in the control station (2400 μmol/kg). The TA drop was especially evident in D2 and D6, with values nearly 400 μmol/ kg lower than that in the Blank station. Moreover, estimated pCO 2 in surface of this reference point was 530 μatm whereas the surface seawater in the sampled sites had pCO 2 values in the range of 595 to 2638 atm, indicating a strong gas oversaturation with respect to the atmospheric CO 2 level in the proximities of the lava cascade. In addition, dissolved organic carbon (DOC) levels in the transect D1D2, corresponding to the active front of the lava flow, were much lower (nearly half) than those obtained in the rest of stations and in the control station, which showed very similar concentrations. The lowest DOC values found could be attributed to instant and rapid thermal degradation in the proximities of the lava front. Similarly, levels of total dissolved nitrogen (TDN) were higher in the affected zone compared to those measured in the remote station. This trend could be linked to the inorganic nutrient enrichment normally associated to volcanic emissions that rapidly introduce nitrogen compounds into the water column although this assumption remains to be tested (Hawkes et al., 2015). Fig. 5. Topographic models generated with the SfM workflow followed for the Condor hexacopter flight performed on October 2nd 2021 at constant 300 m altitude over the lava delta: a) thermal orthomosaic (relative temperature); b) false color composite (NIR-green-red bands); c) NDVI index; and d) SVM classification. The QGIS (v.3.16.14, https://qgis.org/downloads/) and SAGA GIS (v.7.9.0, https://saga-gis.sourceforge.io/en/index.html) programmes were used to create the figures. A. Román et al. Science of the Total Environment 843 (2022) 157092 8 Fig. 6. Vertical profiles of temperature (red line) and salinity (blue line) from the surface to 7 m depth in each sampling station (D1 to D6). A. Román et al. Science of the Total Environment 843 (2022) 157092 9