*4.1. Calibration by Means of MODIS Sea Surface Temperature*

To control the radiometric quality of the TET-1 data, a scale factor was applied as described in Section 2.2.2. By empirically testing different scale factors, the optimal scale factor was derived by minimizing the difference between the atmospherically corrected TET-1 and the MODIS SST. We assume stable temperatures during the short time differences of less than 26 min between the TET-1 and the corresponding MODIS acquisitions.

In the ideal case individual optimal scale factors should have been obtained from and applied to all single TET-1 acquisitions. However, for only two of the nine TET-1 acquisitions an area over the sea was free of clouds. Thus, optimal scale factors could be obtained only for these two TET-1 acquisitions. The optimal scale factor of the 25 April 2015, 17:49 UTC day time scene was applied to all other day time acquisitions. The optimal scale factor of the 21 March 2015, 04:36 UTC night time image was applied to all other night time scenes. Nevertheless, the Tables 3 and 4 show a very small derivation of the temperature differences between the two optimal scale factors for the day and night time scene. In addition, we see that even in the worst case of scale factor 1.00, meaning no 'correction' of the radiometry, the maximum difference between the TET-1 and the MODIS SST was 5K. Therefore, we assume that reasonable scale factors were applied to all nine TET-1 acquisitions analyzed in this study.

#### *4.2. Comparison with MODIS and VIIRS Hotspots*

Table 5 showed the integrated radiant power for each TET-1 acquisition and also the one of the MODIS and VIIRS scenes acquired on the same dates for comparison. There are different reasons for a "missed" hotspot by MODIS or VIIRS, while thermal activity could be detected by TET-1. First, the volcano might be covered by clouds or a volcanic ash plume during an overfly of MODIS or VIIRS. Second, TET-1 is more sensitive for detection of thermal activity than the two other sensors, especially MODIS. Third, a different volcanic thermal activity can be assumed at the different acquisition times of the three sensors.

The following discusses the single acquisitions in more detail. The radiant power of the 22 February 2015, 17:23 UTC and the 27 March 2015 TET-1 acquisitions were not comparable with the corresponding radiant power derived from MODIS or VIIRS data, since the TET-1 acquisitions were taken more than 11 h later. However, the MODIS and VIIRS hotspots confirm the activity of Villarrica Volcano at these two dates.

The radiant power derived from the 9 March 2015, 04:32 UTC TET-1 acquisition is in the same order of magnitude as the one derived from the corresponding VIIRS acquisition (05:57 UTC). In contrast to this, for the 12 March 2015, 04:35 UTC TET-1 scene there is a stronger difference with the corresponding VIIRS scene which was acquired 26 min later. However, both sensors, TET-1 and VIIRS, confirm the thermal activity of Villarrica Volcano for that date.

On 21 March 2015, no hotspot was detected by TET-1. This matches the observations by VIIRS and MODIS. The TET-1 25 April 2015, 04:35 UTC shows a higher radiant power than the one derived from the VIIRS scene acquired 4 min earlier (04:31). However, a second VIIRS acquisition taken 1.5 h later shows a strong increase of the radiant power compared to the first VIIRS acquisition of that date. The radiant power measured by TET-1 is in the middle between these two VIIRS acquisitions. Finally, the high thermal activity detected at the 25 April 2015, 17:49 UTC TET-1 acquisition was confirmed by

the MODIS scene acquired 23 min later. However, the radiant power detected by TET-1 is more than twice as high as the one measured by MODIS.

Overall, these aspects show the major differences of the obtained results, due to the different resolution (and sensitivity), and thus underlines the potential of high resolution thermal infrared sensor systems for this kind of investigation.

#### *4.3. Comparison with Independent Observations*

Moussallam et al. [1] reported that the lava lake at the Villarrica summit crater briefly disappeared on 25 February 2015 and reappeared at the surface on 28 March 2015. The analysis of the TET-1 imagery showed lower radiant power values after the 3 March 2015 eruption than at the TET-1 scene 22 February 2015, three days before the disappearance of the crater lava lake at the surface (cf. Section 3.2). Higher radiant power was again observed on the 25 April 2015, 17:49 UTC acquisition. The subpixel temperatures already showed for the 27 March 2015 TET-1 scene, i.e., one day before the reappearance of the crater lava lake at the surface, and the two following TET-1 acquisitions on 25 April 2015, higher values than before.

#### *4.4. Influence of the Off-nadir Angle and the Depth/Width of the Crater*

As mentioned in Section 1.1, the lava lake at the funnel shaped summit crater of Villarrica Volcano is approximately 20 m to 30 m wide and located at depths of 50 m to over 150 m [15]. As the crater lava lake disappeared on 25 February 2015 (cf. Section 4.3) [1], these conditions are valid for the time before the 3 March 2015 eruption, i.e. for the 22 February 2015 TET-1 acquisition. The off-nadir angle of the center line of this TET-1 acquisition, where Villarrica Volcano is located, is 19.4◦. The deeper the location of the crater lava lake, the smaller is the percentage of the crater lava lake which is visible for the TET-1 sensor. When assuming a crater lava lake width of 30 m, the maximum depth of the crater lava lake (after which the full crater lava lake is in the shadow and not directly visible anymore for the TET-1 sensor under the aforementioned off-nadir angle) is approximately 90 m (maximum 60 m depth, for a width of 20 m). Consequently, we can assume that a part of the crater lava lake was covered by shadow and not visible for the TET-1 sensor. Therefore, there is a high probability that the values of the radiant power presented in Section 3.2 were underestimated. Except for the aforementioned disappearance of the crater lava lake on 25 February 2015 and reappearance at the surface on 28 March 2015 [1], no further information about the depth of the crater lava lake was available for the time after the 3 March 2015 eruption. Nevertheless, we can also assume an influence of the depth of the crater lava lake on the measured radiant power at the other TET-1 acquisitions. Besides the depths of the crater lava lake, the type of volcanic activity (spanning from the lava lake to Strombolian) also influences the detectability from spaceborne sensors [43].

#### *4.5. Detection of Surface Changes*

Regarding the spatial resolution of the thermal channels and the repetition rate, the TET-1 satellite is in between the class of the low spatial resolution/high repetition rate sensors, such as MODIS and Sentinel-3, and the class of high spatial resolution/lower repetition rate sensors such as Landsat-8. Its high flexibility and the aforementioned characteristics make the TET-1 satellite well-suited for rapid detection of larger changes at the Earth's surface. Section 3.3 showed the application of TET-1 data for the detection of the ash coverage of the Villarrica glacier caused by the 3 March 2015 eruption. This flexibility allowed TET-1 to acquire the first image after the 3 March 2015 eruption earlier than the first available post-eruption Landsat-8 acquisition.

We found that the difference of the pre- and post-event MWIR surface temperature is much better suited for the detection of changes in the glacier area than the difference of the reflectivity of the preand post-event RED bands. The best results could be obtained by a combined analysis of both channels.

#### **5. Conclusions**

Villarrica Volcano is, with over 50 eruptions reported since the 16th century, one of the most active volcanoes of the South Andes Volcanic Zone. A time series of nine thermal images of the first satellite Technology Experiment Carrier-1 (TET-1) of the German Aerospace Center's (DLR) FireBIRD mission was analyzed to study the time before and after the 3 March 2015 eruption. We presented an atmospheric correction of the TET-1 data. The emissivity information was derived from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) Global Emissivity Database and the corresponding water vapor data from the Moderate Resolution Imaging Spectroradiometer (MODIS) acquisition with the shortest temporal baseline to the TET-1 acquisitions.

The detected thermal anomalies were investigated at subpixel level by deriving the subpixel temperature, the percentage area coverage of a pixel, and the radiant power. These observations were compared with hotspot information derived from MODIS and Visible Infrared Imaging Radiometer Suite (VIIRS) data. Analysis of TET-1 data showed thermal activity of Villarrica Volcano nine days before the 3 March 2015 eruption. The measured radiant power showed a decrease after this first eruption. An increase of the volcanic activity was again observed on 25 April 2015.

In addition to the analysis of the thermal hotspots at subpixel level, also the eruption-related ash coverage of the glacier at Villarrica Volcano was investigated by means of TET-1 imagery. The changes detected using TET-1 imagery matched well with the reference information derived from higher spatial resolution Landsat-8 imagery.

In summary, the information extracted from the thermal data of the flexible FireBIRD mission could be used in future to support and complement ground-based observations of active volcanoes.

**Author Contributions:** S.P. designed the experiments, performed the data analysis, and wrote the paper. R.R. supported the atmospheric correction of the TET-1 imagery. C.F. supported the TET-1 data analysis. M.N., S.M., T.R., E.S., and D.K. supported the study and provided suggestions for its improvement.

**Funding:** This research was funded, in part, by the German Federal Ministry of Education and Research (BMBF) under grant no. 03G0876 (project RIESGOS).

**Acknowledgments:** The TanDEM-X DEM data was kindly provided by DLR (Proposal ID: DEM\_GEOL1424). The authors would like to thank the three anonymous reviewers for their very constructive remarks.

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

#### **References**


© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
