*4.4. Structures Near the Vent and the Upper Cone Area*

The persistent activity of Fuego volcano, including background activity, causes frequent and often continuous variations in the morphology of the summit. Tracking such changes, which may provide important indications of future activity, is very challenging from the ground and typically requires overflights by manned or unmanned aircraft, which may be infrequent. Figure 8 shows some of these changes identified using PS imagery collected over a period of ~ 4 months, from December 2017 to March 2018. A roughly circular vent in December 2017 evolved into a well-defined circle with a clear summit crater in January 2018. During the paroxysmal activity at the end of January and beginning of February 2018 the crater morphology changes substantially and material appears to overspill down the upper west flank. In subsequent images, cycles of crater excavation and infilling can be discerned (Figure 8). The high temporal and spatial resolution of PS imagery thus provides a novel ability to track morphological variations in the inaccessible regions of active volcanoes, such as Fuego, before, during and after eruptions.

**Figure 8.** *Cont*.

**Figure 8.** PlanetScope images of Fuego collected between December 2017 and March 2018, showing the evolution of the vent region of Fuego volcano. Left panels (**a**, **c**, **e**, **g**, **i**, and **k**) show the images without annotations; right panels (**b**, **d**, **f**, **h**, **j**, and **l**) show the mapped morphology and significant changes (orange lines). Figures a-d display the pre-eruption phase, whereas figures e–l represent the morphology right after the event to one month from it.

#### **5. Discussion**

The February 2018 eruption of Fuego lasted for approximately 20 h [24], and a summary of the activity is shown in Figure 2. The activity was captured by a monitoring webcam on the SE-flank of the volcano, located ~6 km from the vent. A video of the webcam images that tracks the full eruption can be accessed at https://youtu.be/JHwKFPTGAQk and https://youtu.be/4jdfmf3j4ww. Two types of activity can be distinguished: Strombolian to violent-Strombolian activity, characterized by lava fountains, lava flows and a small eruptive column developing over the active vent, and a second type of activity generating PDCs. The variety of eruptive processes and products generated during this short eruption presents a unique opportunity to test the capability of PL satellite data to identify and map volcanic deposits quickly after they have been emplaced.

In the case of the lava flows, the visual mapping using the visible bands gives good results. Attempts to use band differences or index-based methods (e.g., NDVI) did not work well. Although fresh (still hot) lava flows can be mapped using thermal (usually TIR and SWIR) bands [15], our results show that mapping lava flows with high resolution (~3 m pixel size) visible images can be a good alternative provided the new lava flows show enough contrast with the background. The Smithsonian Institution Global Volcanism Program (GVP) reported a maximum length of 800 m for the lava flows, but from the mapping, it is clear that the length reaches ~2000 m for the longest branch and ~1500 m for the shorter lobe (Figure 3). This shows that the detailed mapping possible with PS images can significantly improve the estimation of such parameters. The two branches in the lower part of the lava flow are easily identified from the visible bands but closer to the vent region it is necessary to do a more detailed analysis. The use of a threshold on a single band highlights a larger area which comprises spatter, lava flows and tephra. Visual mapping of the February 2018 lava flows allows us defining some important volcanological features and better understand the importance of this eruption in the history of the Fuego volcano. According to [33] the mean thickness of recent lava flows, which are mainly basaltic in composition, is 2 to 4 m, therefore we estimate an approximate total volume of 900,000 m3 of lava was emitted during this eruption (the total area of lava flows is 0.30 km2, with an average thickness of 3 m). Considering the eruption duration (~20 h), we derive an approximate effusion rate of 12 m3/s. The total volume of lava flows and PDCs (the latter being more difficult to assess, as discussed below) confirms a VEI of 2 [24] for the studied eruption. According to [33], the lava flow length, volume and effusion rate of the February 2018 activity are values typical of paroxysmal eruptions of Fuego. It is worth noting that this paroxysmal eruption occurred only five months before a VEI 3 eruption (June 2018).

Similar considerations apply to the mapping of PDC deposits. Our attempt to map the PDCs using single bands or NDVI differences did not give good results, similar to the lava flows. For smaller eruptions of Fuego, such as in February 2018, the material is confined to the channels on the flanks of the edifice, which, for the most active barrancas, are usually non-vegetated, making the use of change detection techniques based on NDVI difference methods more difficult or impossible. A larger event, such as the June 2018 eruption (VEI 3), produces a larger volume of deposits that may overspill the barrancas or cover less frequently impacted areas downstream, so that the detection can be done not only visually but also based on change detection methods. In the February 2018 case, visual mapping based on visible band combinations worked well.

For the tephra fall mapping analysis the NDVI difference technique gave better results, and we consider this a much better alternative to visual mapping of the deposit, given the gradational boundaries of the tephra fall deposits, i.e., the transition from areas with heavy tephra fall to areas with no tephra fall is smooth and may be difficult to define visually. We highlight that this technique can be used with other sensors that include the necessary bands (visible and NIR). Figure 5 shows a comparison between the results obtained from the analysis performed on PL and Sentinel 2 images, and the results show a good agreement, however the practical implications of higher spatial resolution (~3 m PS pixel compared with the 10 m Sentinel-2 pixel) and temporal cadence (up to daily for PS compared with ca. five days for Sentinel 2) can be significant.

Deposition of volcanic material can change the reflectance of the land surface, depending on the type of surface cover present before the volcanic material is deposited, and on the type of volcanic deposit. If the new volcanic material has very different reflectance properties from the pre-existing surface material, the deposition of the new material will result in a strong reflectance change, and possibly also a strong reflectance contrast with the surrounding areas. For volcanoes in highly vegetated regions (a majority of the volcanoes in tropical and temperate regions), like Fuego volcano, the new volcanic material will often deposit on top of previously vegetated terrain which will result in a strong reflectance change, as volcanic material and vegetation have very different reflectance spectra. We exploit these characteristics by using a NDVI difference technique to detect changes in vegetated areas caused by new deposited material. Although the technique does not work well for non-vegetated areas, the detection of new deposits in such areas is often very difficult, in general, because the new volcanic deposits can have virtually the same reflectance as the pre-existing material and, therefore, a general spectral technique with a limited number of bands may not be able to detect those changes. In such cases a visual inspection may be a quick, simple and effective alternative. Our goal, then, is to focus on the deposition of material in vegetated areas, where we know the NDVI difference method has a high chance of success, as is the case for the tephra fall deposit, described in this paper. For other types of deposits in other areas we use a visual inspection approach, to illustrate that the images can still be useful when analyzed in that way.

It is quite apparent that the NDVI difference results have a sharp boundary near the vent area when showing the mapped tephra fall region (Figure 4). This is most likely due to the presence of ash in the air for the pre-eruption image (from 9 January 2018), as Fuego is a very active volcano. This illustrates that care must always be taken to interpret the data and, in this case, recognize such artefacts.

Finally, the higher spatial resolution of the PS images permits detailed mapping of structural and morphological changes associated with the volcanic activity. The scar at the head of Barranca Honda (Figure 7) is an example of the detail that PS data provides. This structure is interpreted as either a collapse or erosion feature associated with the generation or transit of PDCs through that area, and the identification of this structure in the image acquired immediately after the eruption, puts its possible origin (and the corresponding interpretation) in the necessary volcanological context.

The comparison between PS and other satellite data sources underlines the strength of the PL products. In addition to high spatial resolution, the high temporal cadence permits detailed monitoring and offers a higher probability of obtaining useful data (e.g., Figure 5). The ability to monitor the morphological evolution of otherwise inaccessible volcanic vents is a novel feature of PS images (e.g., Figure 8). A similar analysis to that shown here for Fuego could be applied to volcanoes where lava dome growth or the accumulation of potentially unstable material on the upper flanks constitutes a potential hazard (e.g., through the generation of PDCs) or, in general, where deposited material could be later mobilized, producing lahars. Such analysis could reduce the potential exposure of field-based scientists to volcanic hazards by highlighting potential threats and providing an alternative to field-based deposit mapping. Access and processing of PS data can be performed relatively quickly; locating, downloading, preprocessing, visual inspection and quantitative analysis (e.g., NDVI differences) can take < 1 h to a few hours for a trained analyst. Rapid assessment of volcanic hazards is particularly important in highly populated areas [40], where timely monitoring of volcanic activity (e.g., deposition of volcanic products) and frequently updated products are necessary for hazard mitigation.

#### **6. Conclusions**

Persistently active volcanoes such as Fuego provide an excellent test of PS images as a resource for mapping relatively small and ephemeral, yet significant, volcanic deposits. The paroxysmal activity of February 2018 produced a range of deposits with variable spatial and spectral characteristics which could be mapped using visual analysis and change detection techniques based on single bands or the NDVI. We have shown how such deposit mapping yields estimates of eruption volume and magnitude, with important implications for subsequent activity. The high spatial resolution and temporal cadence of PS imagery also permits the identification of new deposits and morphological changes on scales of a few meters, which are a poorly understood aspect of frequently active volcanoes. These characteristics also proved to be a valuable tool for monitoring pre-eruptive and eruptive phases, which is often impossible from the ground. Similar methods could be applied to other active volcanoes

during volcanic crises. We conclude that PS images represent a valuable new resource for volcano monitoring that promises to increase our understanding of, and ability to observe, volcanic processes.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-4292/11/18/2151/s1, Table S1: Planet Labs imagery timeline data entry for Figure 2.

**Author Contributions:** Conceptualization: A.A., S.C. and R.E.-W.; formal analysis: A.A. and R.E.-W.; funding acquisition: S.C. and G.G.; investigation: A.A.; methodology: S.C. and R.E.-W.; resources: S.C.; supervision: S.C., R.E.-W. and G.G.; validation: R.E.-W.; visualization: A.A.; writing—original draft, A.A.; writing—review and editing: A.A., S.C., R.E.-W. and G.G.

**Funding:** This research received no external funding.

**Acknowledgments:** We thank Planet Labs Inc. for providing an exciting new source of Earth observation data. We also acknowledge the International Geological Masters in Volcanology and Geotechniques (INVOGE) program, which promoted international cooperation between the University of Milano Bicocca and Michigan Technological University. The comments of three anonymous reviewers significantly improved the paper.

**Conflicts of Interest:** The authors declare no conflict of interest.
