the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 29 Jan 2026
Water Vapor Detection in Volcanic Plumes: Near-Infrared Cameras applications at Lascar, Chile, and Litli Hrútur, Iceland
Abstract. Water vapor (H2O) dominates volcanic gas emissions globally with > 70 mol% of total volatile discharge, yet accurate H2O flux measurements remain challenging due to high atmospheric background and H2O’s spectroscopic and physical complexities. We developed a multi-band near-infrared (NIR) camera system, calibrated with an in-situ Multi-GAS instrument to quantify volcanic H2O flux. By combining plume speed measurement with H2O absorption data, we derived the H2O fluxes under favorable atmospheric conditions. We tested our approach at two contrasting volcanic settings: the passively degassing, high altitude, arid atmosphere Lascar volcano (Chile), and at the actively erupting, sea-level, humid atmosphere Fagradalsfjall volcano (Iceland) during the Litli Hrútur 2023 eruption. In November 26–29, 2022, and December 29, 2024, Lascar emitted 23,115 ± 10,694 t d-1 and 46,891 ± 18,863 t d-1 of H2O, respectively, higher than previous estimates using traditional SO2-based methods. In July/August 2023, the Litli Hrútur eruption averaged 19,108 ± 7,560 t d-1 of H2O emissions, matching petrological estimates and steadily declining towards the end of the eruption. The simultaneous deployment of NIR camera, miniDOAS, and a UV camera prove that H2O and SO2 emissions vary independently, with Multi-GAS H2O/SO2 ratios fluctuating over time. This variability challenges traditional measurements and demonstrates that independent direct measurements of major gases (H2O, CO2, and SO2) are essential for accurate volatile budgets and understanding volcanic degassing processes. Our work shows that the NIR camera approach provides a high-rate near-real time and direct method to obtain and visualize H2O emission rates.
- Preprint
(3907 KB) - Metadata XML
-
Supplement
(187 KB) - BibTeX
- EndNote
Status: open (until 30 Mar 2026)
- RC1: 'Comment on egusphere-2025-6430', Christoph Kern, 26 Feb 2026 reply
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 155 | 99 | 15 | 269 | 24 | 9 | 19 |
- HTML: 155
- PDF: 99
- XML: 15
- Total: 269
- Supplement: 24
- BibTeX: 9
- EndNote: 19
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of
Water Vapor Detection in Volcanic Plumes: Near-Infrared Cameras applications at Lascar, Chile, and Litli Hrútur, Iceland
By Felipe Rojas Vilches, Tobias P. Fischer, Scott Nowicki, Melissa Anne Pfeffer, Felipe Aguilera, Tom D. Pering, Thomas Wilkes, Susana Layana, Cristóbal González, Matthew Fricke, John Ericksen, Melanie Moses.
This manuscript discusses efforts by the authors to use cameras recording scattered skylight in the near infrared (NIR) spectral region (850 to 950 nm) to quantify the emission rate of water vapor coming from volcanic vents at Lascar Volcano, Chile, and Litli Hrútur Volcano, Iceland. The first section of the manuscript describes the methods by which the measurements are made and how the collected data are calibrated. Besides the novel NIR method, the authors also employed several established instrument types to obtain information on the emission rate and chemical composition of volcanic gases emitted from the studied volcanic vents. Multi-GAS, DOAS, and UV Camera systems were deployed at both sites, and the measurements collected with these instruments are compared to those from the NIR camera system and, in the case of the Multi-GAS, used for calibration purposes. Finally, the obtained measurement results are summarized, interpreted with regards to their implications for volcanic activity occurring at the examined volcanic systems, put into context of the existing literature, and compared to other regional and global studies.
An impressive amount of work went into the development of this study as the authors not only describe a new measurement technique but also outline attempts at comprehensive characterization of activity occurring in two very different volcanic settings with a whole suite of instrumentation. Many of the data, interpretations, and comparisons presented here represent valuable contributions to the community of researchers interested in learning about the examined volcanic systems and the varied manifestations of volcanic degassing mechanisms that they represent.
However, I do have substantial concerns about the water vapor imaging method applied by the authors to quantify water vapor column densities and emission rates. Because many of the datapoints presented in the second half of the manuscript are derived directly or indirectly from the NIR water vapor imaging measurements, and because this method is not an established monitoring technique, the authors must first provide evidence that their instrument is indeed operating as desired and capable of making the measurements for which it is being utilized. Atmospheric Measurement Techniques is indeed an ideal platform for discussing this methodology. In the following, I would like to initiate this discussion, at this stage focusing only on the technical and physical aspects of the imaging method for determining water vapor column densities. Some additional comments on later sections of the manuscript can be found in the annotated document attached to this review, but I recommend that the editorial focus first be placed on the discussion of the water vapor measurement technique, as may of the later results build on these measurements.
1. Multispectral trace gas imaging theory
Generally, the water vapor imaging approach taken by the authors follows an approach commonly used in ultraviolet (UV) measurements of sulfur dioxide (SO2). The so-called SO2 camera combines images taken in two narrowband UV channels to derive the optical depth of SO2 in volcanic gas plumes (Mori and Burton, 2006). The basic idea is as follows: According to the Beer-Lambert-Bouguer Law of absorption, the intensity I of a beam of light passing through an SO2 cloud is reduced as a function of the SO2 concentration c, the path length L in the cloud, and the SO2 absorption cross-section σ.
I = I0 · e-σcl (eq. 1)
Note that σ is a function of wavelength, so to see a marked decrease in intensity, measurements must be performed at a wavelength at which σ is large enough such that the product σ·c·L (hereafter referred to as the optical depth τ) is sufficiently large to impact I. This is achieved by placing a narrowband filter in front of the camera system with its maximum transmittance aligned with an SO2 absorption feature. In the case of the SO2 camera, that ‘on-band’ channel is typically placed near 310 nm. For the water vapor measurements, the authors propose an equivalent on-band channel be placed at 940 nm.
Remote sensing measurements such as these cannot resolve the concentration of a gas species (a point that should be clarified at several places in the manuscript; see in-line comments), but instead measure the integral of the concentration along the light path. This is commonly referred to as the column density S.
S = Integral(c·dl) = cavg · L (eq. 2)
Note that the integral can also be expressed as the product of the average concentration cavg times the light path length L which, in the case of a gas plume, could be the plume width. Given a measured column density S, it is possible to use this relationship to estimate an average concentration if the plume width is known.
Solving eq. 1 for the column density yields:
S = σ-1 · ln(I0/I) (eq. 3)
Since σ is known (it can be measured in the laboratory and, in the case of water vapor, is available in the HITRAN database), S can be determined from high-spectral-resolution measurements of I and I0. Or, as is the case for the low-resolution multispectral imaging described here, S is approximately proportional to ln(I0/I), with the proportionality constant derived in a calibration process. Again, I is the intensity of light arriving from the volcanic plume, and I0 can typically be estimated by assuming that the incident radiance directly adjacent to the plume is the same as it would be at the location of the plume if the absorbing gas were not present.
A complicating factor with this method is that trace gas absorption is not the only process influencing the intensity of light incident from the volcanic plume. This is because volcanic plumes often contain aerosols and water droplets which can scatter light towards the camera system. For this reason, plumes often appear brighter (and white in the visible spectrum) than the surrounding sky. In such cases, simply applying eq. 3 to determine the column density can yield negative values for S (and c).
Following a methodology originally developed for measurements of SO2 emitted from industrial stacks (McElhoe and Conner, 1986), Mori and Burton (2006) solved this issue by introducing an off-band reference channel to their imaging system. The idea is that this off-band channel should be outside of the absorption range of the trace gas of interest but close enough such that atmospheric scattering processes are approximately the same in both channels. This then allows the measurement of ln(I0/I) to be normalized by the radiance expected in the absence of the absorbing trace gas. This is the standard definition of differential optical depth for multispectral camera systems, also known as the “apparent absorbance” (see Kern, 2025, for additional details). The authors provide this same definition in their eq. 5. The idea here is that the apparent absorbance should be roughly proportional to the water vapor column density along the path of incident radiation. This then allows the constant of proportionality to be determined by calibration, e.g. using a Multi-GAS measuring the water vapor concentration in the plume.
2. Image processing
In principle, the very same approach that is commonly used by SO2 cameras in the UV can also be used in NIR imaging systems utilizing images at 940 nm (on-band) and 850 nm (off-band). However, the authors introduce an additional processing step. Beginning in line 237 of the manuscript, several sentences describe use of visible imagery to attempt an additional “brightness correction”. The exact process suggested by the authors is a bit difficult to follow without explicit equations defining the processing steps, but it appears that an image of incident radiance measured in the visible spectral region is subtracted from each NIR image (both on- and off-band channels).
It is unclear to me what the purpose of this step is, and I have serious concerns about the validity of such a “correction”. For one, an image recorded for human perception in the visible region would be sufficiently far from the NIR channels of the measurement that the assumption of scattering effects being identical at the two considered wavelengths would fail. Volcanic plume aerosols tend to be very small and can have Angstrom exponents of up to about 2.5, meaning that the aerosol optical depth at 500 nm could be up to 4 times greater than that at 900 nm (see Kern, 2025). Therefore, scattering processes would be quite different at the two wavelengths. Also, since the visible image stems from a different sensor, it is unclear how a subtraction operation would not invalidate the quantitative nature of the NIR measurements. The visible camera system would certainly have a very different light throughput and quantum efficiency as the NIR cameras with bandpass filters placed in front of them.
Finally, it is also unclear how subtraction of a brightness image (measured at any wavelength) would provide a valid correction for scattering processes. Instead, I believe it invalidates the assumed roughly linear relationship between measured optical depth and water vapor column density. Imagine a measurement in which a high cirrus cloud leads to visibly increased brightness in part of the image, and assume this cloud is above the vast majority of atmospheric water vapor (cirrus clouds are commonly above 10,000 m). At the location of the cloud in the image, such a correction k would reduce the NIR intensity in both the on- and off-band channels, thus adjusting the optical depth from τ = ln(I0/I) to τ = ln(I0-k/I-k). This would change the estimate of water vapor derived from this measurement, even though the incident radiation is still passing along the same path through the same water vapor layer much lower in the atmosphere. In summary, such a correction seems invalid and unnecessary, as the normalization by the off-band channel occurring within the calculation of apparent absorbance is already meant to correct for a change in brightness due to scattering effects (although scattering causes additional problems when measuring water vapor; see next section).
3. Correct quantification of water vapor in the atmospheric background
As mentioned above, I believe that the brightness authors’ brightness correction should be eliminated to maintain a roughly proportional relationship between apparent absorbance and water vapor column density. In that case, the described methodology would mirror that of SO2 cameras operating in the UV spectral region and, in principle, the NIR imaging method should yield apparent absorbance images in which the apparent absorbance is related to the water vapor column density along the light path of the incident radiation. However, there is a fundamental difference in measuring water vapor vs SO2. While SO2 is only present in volcanic plumes and the concentration is negligible in the background atmosphere, this is not the case for water vapor. As the authors point out, the background atmosphere contains significant amounts of water vapor, so the challenge here is whether the additional water vapor in a volcanic plume can be detected above the atmospheric background.
This problem is more complex than it may seem at first. The authors argue that the concentration of water vapor could be orders of magnitude higher in the volcanic plume than in the background, particularly if the temperature of the plume is tens of degrees Celsius higher than the background air temperature. This is convincingly demonstrated by calculating the water saturation vapor pressure as a function of temperature. However, it is important to remember that the remote sensing instrument is sensitive to the column density of water vapor along the light path, not the concentration itself. This has several implications.
Water vapor is not evenly distributed in the atmosphere. Because the water vapor pressure increases exponentially with temperature (Clausius-Clapeyron relation), the atmosphere’s ability to hold water in vapor state is highly dependent on temperature, and therefore on altitude (see Kern et al., 2017). In fact, the scale height of water vapor in the atmosphere is only about 2 km, while the scale height of air (mostly nitrogen and oxygen) is about 8 km. This means that most water vapor in the overhead atmosphere is located in a relatively thin overhead layer, no matter where an instrument is located on Earth. For this reason, and because the scale height of water vapor is so much smaller than that of air, it is reasonable to assume that skylight scattered towards an instrument was last scattered above the water vapor layer and has passed straight through it along the instrument line of sight. More details on this “geometric approximation of air mass factor” are given in Kern et al. (2017).
Under this assumption, scattered skylight passing vertically through the atmosphere will pass through the water vapor layer along a straight vertical line, and the integral of the water vapor concentration along this line is the vertical column density (VCD). At sea level, the average global water vapor VCD is approximately 1·1023 molecules/cm2, although actual VCDs vary substantially with latitude/temperature. A volcanic plume containing 2·1018 molecules/cm3 water vapor with a thickness of 500 m would exhibit the equivalent column density which, according to the authors’ eq. 1, requires a plume temperature above about 43 C. Regardless of the exact assumptions and numbers, the point is that measured atmospheric water vapor column densities are expected to be at least of the same order of magnitude as those contributed by a volcanic plume, especially if the measurements are made at low elevation angles rather than aimed straight at the zenith.
A reasonable test of the instrument and methodology is therefore whether variations in atmospheric water vapor column densities can be imaged. An ideal experiment in this regard is to look for the expected variation in water vapor column density as a function of elevation angle in images of the background atmosphere. Assuming the geometric approximation of air mass factor, the background water vapor column density should scale with 1/sin(α), where α is the elevation angle above the horizon. Although not explicitly given in the manuscript, the Logitech C300 likely has a field of view of several 10s of degrees. If aimed just above the horizon on a clear, cloud-free day, the difference between water vapor column densities measured at the bottom of the image and those measured near the top of the image should be significant and easily quantifiable, especially for low-altitude sites. But even considering the measurements performed by the authors in Chile, this 1/sin(α) relationship should be clearly visible in the imagery. Based on the description of measurements at Lascar from section 3 of the manuscript, the images shown in Figure 8 likely span a range of elevation angles of at least 10 to 20 degrees above the horizon. According to the geometric approximation of air mass factor, the background water vapor column density measured at 10 degrees elevation S10 should be 5.8 times the VCD, while the background column density S20 measured at 20 degrees should be about 2.9 times the VCD. This means the background column density (measured in a plume-free area) should vary by about a factor of two vertically across the image. If the field of view is greater than 10 degrees, that variation should be even larger. The authors don’t include clear-sky images in their current manuscript, but figure 8B contains what appears to be a mostly plume-free area on the left side of the image. However, there is no discernable difference between the apparent absorbance measured at the bottom vs the top of the plume-free area, which is quite concerning (although this might be due to an invalid brightness correction; see previous section). The situation would be even more severe in Iceland due to the higher water vapor VCD, although I’m not sure if cloud-free data are available here. In summary, any system that relies on detecting the absorption of radiation by in-plume water vapor must also be able to clearly measure the expected 1/sin(α) progression of background water vapor column density (compare to Kern et al., (2017), figures 7 and 9). This is not convincingly shown in the manuscript as it stands.
4. Scattering effects
In sections 1 and 2 above, the calculation of apparent absorbance was discussed and how normalization of SO2 camera measurements using an off-band optical-depth can be used to correct for the effect of scattering on in-plume aerosols and water droplets. The idea is that any changes in image brightness related to scattering processes are captured by the off-band images, and normalization by these images thus corrects for changes in plume brightness not related to SO2 absorption. Notably, however, this correction method does not correct for any changes in the light path distribution that radiation is taking on its way from the sun to the instrument. In the case of SO2, such changes typically have a minor effect on the measurement results unless plumes become severely opaque or optically thick (Kern et al., 2013). This is because the SO2 measurements are largely indifferent to the path that radiation takes through the atmosphere as long as most measured light passes through the plume along a straight line towards the instrument.
This is not true for water vapor measurements. Because the background atmosphere contains a significant amount of water vapor, the effective path that radiation takes on its way from the Sun to the instrument will determine the magnitude of the measured background signal. As discussed in the previous section, the path of radiation through the atmospheric water vapor background is about 5.8 times longer than the straight vertical line if an instrument is aimed 10 degrees above the horizon on a clear day. If, however, scattering aerosols or water droplets are introduced to a location near the instrument, radiation scattered on these is no longer subject to the geometric approximation of air mass factor and can pass through the atmospheric water vapor layer along much shorter (or longer) paths. In one common model, the “single-scattering-approximation”, all detected radiation is scattered only once in the atmosphere. This model is valid in cloud-free atmospheres or those that only contain optically thin clouds or plumes. In this approximation, radiation scattered on plume aerosols would originate from the sun, fall straight on the plume, then pass straight on to the instrument. Especially if the sun is high overhead (near solar noon), the resulting light path through water vapor in the atmospheric background would be significantly shorter than when looking through the atmosphere at a low elevation angle. In such conditions (clear sky, low solar zenith angle, lightly scattering plumes), this should actually result in negative water vapor optical depths measured in plumes (also see Kern, 2017). In the authors’ Figure 8a, there appear to be a few places where such ‘negative plume’ anomalies might be visible, although it is unclear if they could be due to other effects of the applied image processing methods (see discussion in section 2 above). Although such negative plume anomalies would prevent any reliable estimate of in-plume water vapor concentration, they would at least show that the physics of the measurement are being correctly recorded in the measurements.
In summary, any scattering that occurs on aerosols or water vapor droplets in the plume lead to significant changes in the atmospheric light path which, in the case of water vapor, will lead to a breaking down of the ability to separate the in-plume water vapor signal from that of the background. Scattered light measurements of in-plume water vapor based on spectral absorption can therefore only succeed in the complete absence of aerosols or water droplets in the volcanic plume, and in the absence of any low to mid-level clouds. Based on the visible images included in the manuscript, these conditions do not appear to have been met, at least during the measurement examples shown (even Figure 14 B shows some scattering in the Lascar plume, although maybe not at greater distance from the vent).
5. Origin of the signals observed in NIR imagery
Given the presented methods and images, it is difficult to ascertain the exact origins of the observed plume signals in the imagery. Plumes are discernable in many of the processed images presented in Figures 8, 11, and 12, and in the supplemental videos. As discussed in section 2 above, the authors’ brightness correction (subtraction of visible light images from NIR scenes) is physically problematic and can have unintended effects. This may partially explain the signals.
Figure 12G and some of the supplemental videos also appear to show some radial effects, with increased NIR apparent absorbance in the center of the images when compared to the edges. Similar effects have been observed in UV camera systems and are typically caused either by unequal vignetting in the two spectral channels or by a shift in the bandpass filters’ maximum transmittance wavelength when light passes through them at an angle (see Figure 5 in Kern et al., 2010). Positioning the bandpass interference filters between the object lens and the CCD sensor can help reduce the latter of these two effects, if such a design is possible with the utilized cameras (Kern et al., 2010).
The remaining signals could be at least partially due to infrared absorption by water vapor although, as pointed out in section 3 above, one would expect a larger variation vertically across the images stemming from water vapor in the background atmosphere. The plume signal may instead be caused by changes in the effective atmospheric light path of radiation scattered on water droplets and aerosols in the plume (see previous section). This scattering effect has been used by others to estimate water fluxes from volcanic vents. Girona et al. (2015) introduce a framework similar to the one presented here in which the luminance of radiation scattered on water droplets in volcanic plumes is related to their water vapor content. Girona et al. (2015) apply their method to laboratory experiments and find a roughly linear relationship between luminance and plume water concentration. To obtain absolute values for water vapor flux from volcanic vents, the brightness values would then need to be calibrated with an independent method, e.g. a Multi-GAS, as suggested by the authors of this manuscript.
I believe that the signals observed in the processed NIR imagery presented in this manuscript are most likely caused by a combination of these effects. This would not necessarily prevent such a system from being deployed to estimate water vapor emissions at volcanoes (see Girona et al., 2015), but the physics behind the measurements must be properly understood in order to evaluate the limitations of the technique. Currently, the authors explain their signals as stemming exclusively from water vapor absorption in the NIR spectral region. I would argue that this is clearly not the case, given the inability of the system to detect the expected 1/sin(α) vertical profile in the background atmosphere. Since the authors are attempting to introduce the NIR measurements as a novel technique for remote sensing of volcanic water vapor emissions, they must provide evidence that the system is indeed capable of quantitatively measuring NIR absorption by water vapor despite all the other effects described above. Successful laboratory experiments could help here (again, see Girona et al., 2015 for an example), as could experiments aimed at measuring the water vapor VCD in the background atmosphere in cloud-free conditions (see Wagner et al., 2013 for some ideas on how this could be achieved).
6. Comparisons with Multi-GAS
Apart from the newly introduced visible brightness correction, which I believe to be problematic (see section 2 above), the NIR imaging method described in this manuscript appears to be identical to the method proposed by Pering et al. (2017). It attempts to apply the methodology used by SO2 cameras operating in the UV to detect water vapor in the NIR. As outlined in Kern (2017) and in the sections above, there are many reasons why water vapor measurements are substantially more difficult than SO2 measurements using this approach, yet most of these difficulties were left unaddressed in this manuscript.
The one challenge that was discussed by the authors is related to the sensitivity of the NIR camera system to water vapor in volcanic plumes. In past calculations aimed at determining the instrument sensitivity, the plume temperature was assumed to be similar to that of the background atmosphere. As the authors point out here, the plume temperature determines the saturation vapor pressure of water and thus determines the maximum water vapor concentration that is possible before water begins to condense into the liquid phase and form droplets. For example, this circumstance was used by Matsushima and Shinohara (2006) to estimate vent exit temperature for visual observations of condensing/non-condensing plumes.
When volcanic plumes are emitted into the atmosphere, they are quickly diluted by orders of magnitude with background air. During this process, their temperature drops drastically towards that of the atmosphere. However, the authors are correct that it is possible in some cases for plumes to maintain temperatures significantly above that of the background for some time. This depends on the initial volcanic gas temperature, the efficiency with which volcanic gases are mixed with background air (driven by plume dynamics), and the release of latent heat as water condenses in the plume (which itself depends on the meteorological conditions). Based on personal experience from other volcanoes, I find it unlikely that the plumes at Lascar maintained temperatures of 25 to 55 degrees C (assumptions in A1) above background in the region in which the images were collected, but it can’t be ruled out.
Fortuitously, the authors were well-prepared to address this question in their measurements. By bringing a UAV Multi-GAS equipped with temperature and relative humidity sensor to the field, they were able to probe plume temperatures not only in close proximity to the degassing vents, but also within the more distal plume where the NIR imaging method was applied. Unfortunately, unless I missed them somehow, these data are not presented or discussed in the manuscript. Section 3.1.1 gives some maximum relative humidities, but not temperatures, and even the humidity values appear to stem from instruments placed on the crater rim rather than flown through the bulk plume. Table 1 gives some maximum temperatures, but it is unclear if these are from the UAV measurements or again from instruments placed near the vents.
If the data exist, I would like to encourage the authors to present temperatures and relative humidity values from the UAV flights. It would be extremely useful to see these plotted as a function of location or distance from the active vents. If possible, it could also be useful to examine the relationship between temperature and SO2 concentration, as the SO2 concentration can often be used as a plume tracer and gives an idea of the level of plume dilution. This, combined with the SO2 camera measurements could shed some light on the plume temperature, if no direct measurements (by UAV) exist.
Generally, many of the reported plume composition ratios have a high uncertainty. It would be instructive to see the raw data, e.g. plots of H2O concentration vs CO2 concentration, and/or H2O concentration as a function of SO2 concentration. This would not only give readers an idea of the quality of the correlations, but also provide information on the minimum and maximum concentrations of all species encountered during the measurements.
7. Recommendations
The above discussion is focused mostly on the introduction of the NIR imaging technique as a new method for measuring water vapor emissions from volcanoes. This is an appropriate topic for a manuscript under consideration for Atmospheric Measurement Techniques (AMT). I have outlined my concerns above and feel that the study would need to be substantially revised before offering convincing evidence that the methodology is sound. Only at that point would it be suitable for further discussion and possibly eventual publication in AMT.
However, the manuscript also contains a large number of measurements along with discussion and interpretation of results utilizing established volcanic gas monitoring techniques. One option for the authors to consider might be to split the manuscript. This would allow the NIR imaging technique to continue to be refined and discussed while at the same time presenting the collected volcanic gas geochemistry data to the volcanology community. In that case, the geochemical data should probably be submitted to a volcanology journal to best reach the target audience. Of course, the final decision is up to the authors.
I would like to thank the authors and editors for the opportunity to review this work and sincerely hope that my comments will help in moving this work forward. Please do not hesitate to contact me in the event of any questions.
Cited References
Kern, C., Kick, F., Lübcke, P., Vogel, L., Wöhrbach, M., Platt, U., 2010. Theoretical description of functionality, applications, and limitations of SO2 cameras for the remote sensing of volcanic plumes. Atmos Meas Tech 3, 733–749. https://doi.org/10.5194/amt-3-733-2010
Kern, C., Werner, C., Elias, T., Sutton, A.J., Lübcke, P., 2013. Applying UV cameras for SO2 detection to distant or optically thick volcanic plumes. Journal of Volcanology and Geothermal Research 262, 80–89. https://doi.org/10.1016/j.jvolgeores.2013.06.009
Kern, C., Masias, P., Apaza, F., Reath, K.A., Platt, U., 2017. Remote measurement of high preeruptive water vapor emissions at Sabancaya volcano by passive differential optical absorption spectroscopy. J Geophys Res Solid Earth 122, 3540–3564. https://doi.org/10.1002/2017JB014020
Kern, C., 2017. The Difficulty of Measuring the Absorption of Scattered Sunlight by H2O and CO2 in Volcanic Plumes: A Comment on Pering et al. “A Novel and Inexpensive Method for Measuring Volcanic Plume Water Fluxes at High Temporal Resolution,” Remote Sens. 2017, 9, 14. Remote Sens (Basel) 9, 534. https://doi.org/10.3390/rs9060534
Kern, C., 2025. Ultraviolet and visible remote sensing of volcanic gas emissions. Journal of Volcanology and Geothermal Research 468, 108423. https://doi.org/10.1016/j.jvolgeores.2025.108423
Matsushima, N., Shinohara, H., 2006. Visible and invisible volcanic plumes. Geophys Res Lett 33, 2–5. https://doi.org/10.1029/2006GL026506
McElhoe, H.B., Conner, W.D., 1986. Remote Measurement of Sulfur Dioxide Emissions Using an Ultraviolet Light Sensitive Video System. J Air Pollut Control Assoc 36, 42–47. https://doi.org/10.1080/00022470.1986.10466043
Mori, T., Burton, M., 2006. The SO2 camera: A simple, fast and cheap method for ground-based imaging of SO2 in volcanic plumes. Geophys Res Lett 33, 1–5. https://doi.org/10.1029/2006GL027916
Pering, T.D., Mcgonigle, A.J.S., Tamburello, G., Aiuppa, A., Bitetto, M., Rubino, C., Wilkes, T.C., 2017. A Novel and Inexpensive Method for Measuring Volcanic Plume Water Fluxes at High Temporal Resolution. Remote Sens (Basel) 9, 1–13. https://doi.org/10.3390/rs9020146
Wagner, T., Andreae, M.O., Beirle, S., Dorner, S., Mies, K., Shaiganfar, R., 2013. MAX-DOAS observations of the total atmospheric water vapour column and comparison with independent observations. Atmos Meas Tech 6, 131–149. https://doi.org/10.5194/amt-6-131-2013