| 11 Aug 2026
A cost-effective, open source, high resolution visible wavelength camera network: design and applications for long-term monitoring and research in volcanology
Abstract. Monitoring and forecasting volcanic processes are crucial for detecting, characterising, and assessing associated hazards. However, advanced monitoring tools, such as thermal cameras, can be expensive and unavailable in many monitoring networks. To address this, a cost-effective visible wavelength camera network has been developed to capture high resolution imagery for volcano monitoring and research.
This network provides continuous, automated image acquisition from different locations around a volcanic edifice. Each camera node is equipped with a Raspberry Pi computer board, one or two Raspberry Pi Camera Module 3 cameras, and a waterproof and dustproof box for deployment in challenging outdoor environments, with an estimated cost of approximately EUR 152–257 per unit, depending on the hardware configuration. The system supports diverse configurations, including adjustable capture intervals and resolutions, to meet various observational needs. The network adapts its operation based on environmental conditions, automatically selecting the shutter speed during daylight and using specific settings at night. Acquired data can be transferred in near real time to a server for archival and, potentially, analysis for monitoring purposes.
In this study, we present a long-term camera monitoring network deployed at Mt. Etna, together with short-term deployments at Stromboli, the Geneva Jet d'Eau as a calibrated experiment, and laboratory experiments. These case studies demonstrate the long-term reliability of the system for capturing images over extended periods under variable lighting and weather conditions, including high temperatures, snow, rain, and tephra fallout, as well as its suitability as a temporary observation tool for field campaigns or laboratory setups. By using open source hardware and software, this camera network provides a flexible and accessible tool for volcano monitoring and research applications. When paired with cellular connectivity and solar panels or battery packs, the system could be integrated into multisensor monitoring networks, including in remote areas and challenging outdoor environments. Although developed for volcano monitoring, the framework can be adapted to other natural and environmental processes requiring persistent or high temporal resolution visual observations.
This manuscript presents a low-cost camera system based on Raspberry Pi hardware for long-term volcano monitoring, temporary field deployments and laboratory experiments. The topic is fully relevant to “Geoscientific Instrumentation, Methods and Data Systems”, particularly through its potential to make visual monitoring more accessible to observatories with limited resources. The flexible design, emphasis on open software and documentation, and demonstrations across several observational settings are valuable contributions. The text itself is well written and easy to read, although some word-choice improvements and minor grammatical or phrasing corrections could sometimes help.
Despite these qualities, the manuscript does not yet sufficiently establish the measurement accuracy and operational reliability needed to support its conclusions. Key aspects require further methodological detail and quantitative assessment, particularly acquisition timing and synchronisation with other instruments, camera calibration and geometric modelling, uncertainty propagation, and power autonomy. The validation should more clearly distinguish agreement between camera-derived measurements from independently established accuracy, while the long-term reliability claims need supporting operational statistics.
These issues require substantive clarification, additional information and, where necessary and possible, further testing. I therefore recommend major revision before the manuscript can be considered suitable for publication.
MAJOR COMMENTS
Here are the seven improvements I consider required to make the manuscript suitable for publication:
1. Defining and validating acquisition timing (Sections 2.2, 2.4, 3.2.2 and 4.1; lines 126–141, 216–225, 280–290 and 349–352)
This is a major missing aspect in the current manuscript. The timing architecture is not properly presented, while it represents a key aspect for a monitoring instrument. First, it should be introduced in Methods and it is not the case. We only discover that in the discussion part. Next, it is not specified whether timestamps represent exposure, acquisition requests or file creation. Such detail is critical for any comparison with other dataset. An assessment of timing offsets, interval variability and clock drift, including after reboots and during disconnected operation, is not provided. The implemented NTP synchronisation method is poorly described and not evaluated. RTC (and why not GNSS) options could be discussed in more details against application-specific requirements, as the current camera system with its current description offers no guaranty of reliability for accurate comparison with other types of measurements, such as seismic and geodetic data. Visual correspondence with visible plume emergence alone does not validate synchronisation.
2. Documenting camera calibration and validating the geometric model (Sections 2.1, 2.4 and 4.2; lines 98–109, 193–215 and 366–370; Eqs. (1)–(5))
Camera’s interior orientation is presented in the discussion as a main source of error for spatial measurements. Intrinsic parameters, lens distortion, focus settings, camera position and orientation, with calibration residuals, should be reported. The authors should also explain how calibration applies to different lenses, acquisition modes and protective windows. They should also discuss calibration stability and establish criteria for recalibration. Equation (1) assumes linear pixel-to-angle mapping, which differs from perspective projection: the authors should justify this approximation quantitatively or replace it with a calibrated model. The authors should validate projected positions using independent control points.
3. Strengthening the independent measurement validation (Sections 2.3.2, 3.2.1 and 4.2; lines 163–171, 264–279 and 372–379)
The authors should specify the Canon reference camera’s calibration, uncertainty and temporal alignment. Agreement between two camera-derived series does not establish absolute accuracy when they may share geometric or processing errors. The authors should supplement the comparison with independently constrained target measurements, or explicitly limit the conclusion to inter-camera agreement. The nominal fountain height should not serve as contemporaneous ground truth.
4. Providing reproducible uncertainty estimates (Sections 2.4, 3.1.1, 3.2.2–3.2.3 and 4.2)
The authors should explain how geometric, timing, target-picking, DEM and plume-direction uncertainties propagate into elevations and velocities. They should document the derivation and statistical meaning of the reported 20–26 m plume-height uncertainty. They should also add appropriate uncertainty intervals to lava-front elevations and laboratory velocities, and assess whether reported changes and peak velocities exceed measurement uncertainty. Figures 5, 7 and 8 should be updated accordingly.
5. Characterising effective high-speed imaging performance (Sections 2.3.3–2.3.4 and 3.2.1–3.2.3; lines 179–190, 273–279 and 292–298)
Authors should report actual frame intervals, dropped frames, exposure durations and sensor modes in a more specific way. They should assess rolling-shutter effects and motion blur for the measured velocities. They should describe differentiation and any smoothing, and test their influence on peak velocity. Nominal frame rate alone does not establish the temporal resolution of a quantitative measurement.
6. Making power requirements and autonomy reproducible (Sections 2.1, 3.2 and 4.1; lines 110–123, 258–262 and 329–345; Table 1)
The authors should expand the Methods section beyond identifying power supplies. They should report measured consumption for each tested configuration and acquisition/transfer mode, usable battery energy and autonomy-test conditions. They should reconcile the broadly stated 20–24 h runtime with the mode-dependent estimates. They should describe behaviour during power interruption and restoration, and distinguish tested autonomous configurations from proposed extensions.
7. Quantifying long-term reliability and data completeness (Sections 2.2, 2.3.1, 3.1 and 4.1; lines 137–141, 150–160, 228–237 and 300–354)
The authors should provide a deployment table showing node configuration, operating duration, expected/acquired images, outages, failures and interventions. They should separate evidence for the original microSD configuration from the SSD replacement. They should document recovery and storage behaviour during connectivity loss. They should distinguish nearby ambient temperatures from measured enclosure conditions, and qualify reliability claims where supporting measurements are unavailable.