the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 17 Mar 2026
Quantifying calibration drift in low-cost PM10 and sound sensors across contrasting indoor environments
Abstract. Low-cost sensor networks offer an affordable option for long-term air quality and noise monitoring, but their performance can be limited by calibration drift. This study quantified drift in low-cost particulate matter (PM) and sound level sensors using measurements from three co-located units and reference instruments deployed in two contrasting indoor environments: a lower-concentration painting environment and a higher-concentration swine environment. Drift varied substantially by environment. In the painting environment, drift in both PM10 and sound sensors was small (PM10: 2.1 μg m-3 per year; sound: 0.78 dBA per year) and either within or near expected performance ranges based on manufacturer specifications and sound level tolerance criteria. In the swine environment, PM10 sensors exhibited rapid degradation (−46.2 μg m-3 per year), indicating substantial time-dependent drift under elevated particulate concentrations. Incorporating a drift term in the calibration model reduced RMSE and MAE by approximately 11 % in the swine environment, with negligible improvements in the painting environment. Sound sensors showed minimal drift and strong inter-sensor agreement across both deployments. These findings highlight the need for environment-specific recalibration strategies for low-cost PM sensors and demonstrate that drift-aware modelling can improve measurement reliability and support maintenance and lifecycle planning in high-concentration settings, with future work investigating how particulate characteristics and operating conditions influence drift.
- Preprint
(6906 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-1248', Simon O'Meara, 14 May 2026
-
AC1: 'Reply on RC1', Oliver Stroh, 22 May 2026
We thank the reviewer for the thoughtful comments regarding journal scope and the inclusion of sound measurements.
We agree that the particulate matter component is more directly aligned with the traditional scope of Atmospheric Measurement Techniques. The intent of including sound measurements was not to position the manuscript as an acoustics-focused study, but rather to examine calibration stability across multiple low-cost environmental sensing modalities deployed within the same monitoring platform.
In response to the reviewer’s comment, we will revise the manuscript to reduce the prominence of the sound-related material and clarify its role as a secondary component intended to provide broader context for the calibration-stability framework. We will also revise portions of the introduction and discussion to place greater emphasis on the particulate matter drift analysis and its implications for environmental monitoring applications relevant to AMT readership.
We appreciate the reviewer’s perspective and believe these planned revisions will improve both the focus and scope alignment of the manuscript.
Citation: https://doi.org/10.5194/egusphere-2026-1248-AC1
-
AC1: 'Reply on RC1', Oliver Stroh, 22 May 2026
-
RC2: 'Comment on egusphere-2026-1248', Anonymous Referee #2, 14 May 2026
Sensor drift with time is a very important factor for the accuracy of low-cost sensors. It is also known that sensor drift is greater on more polluted environments, which makes the idea of the study interesting. There is though a very important aspect in this study which weakens its outcomes, and this is the sensor used. The study employs the SPS-30 sensor which does not measure PM10, but instead it estimates them from measurements of smaller particles and anticipated PM ratios. This in combination with the variable conditions within the swine environment (as the populations of the pigs was reduced substantially over time, and so probably the PM concentrations as well) makes me very skeptical of the outcomes of the study.
The drift of the sensor is presented in bulk and not with time to study its evolution and whether it is sensible or just the difference between measurements in time A and B. The inability of the sensor to measure PM10 is even more obvious in figure 3c in which, despite what is described in the text (chapter 3.3, wrongly titled as "painting environment"), we can clearly see that PM10 is continuously underestimated more than 3 times with variable ranges, regardless of high or low concentrations.
Thus, the drift identified, while plausible and possibly expected, may be greater or smaller by huge margins depending on the particle size distributions (the sensor practically considers specific PM10 to PM2.5 ratios for the PM10 estimation, which is not ideal for a PM10 heavy environment). This also raises the question of why PM2.5 is not presented (which is what the sensor is better suited for)?
Other comments associated with the lack of an intercomparison between the sensors before and after the study (very important to see whether the differences on the drift are real), and that despite all that the Discussion part is well written (though I am not sure if the points made apply for this study) are of no importance, as the study may be well conceived but not properly executed and/or presented.
Citation: https://doi.org/10.5194/egusphere-2026-1248-RC2 -
AC2: 'Reply on RC2', Oliver Stroh, 22 May 2026
We thank the reviewer for the detailed and constructive comments regarding interpretation of the SPS30 PM₁₀ measurements collected in this study.
We agree that the SPS30 does not directly measure PM₁₀ using a size-selective reference method. As described in the manufacturer specification statement, the PM₁₀ output is estimated from optical scattering measurements from smaller particle fractions and assumed aerosol profiles rather than direct measurement of coarse particles. We also agree that this limitation becomes especially important in environments with potentially atypical or variable particle size distributions.
The reviewer additionally raises an important point regarding the reduction in swine population during the deployment period and the possibility that temporal changes in aerosol composition or particle size distribution may contribute to the observed changes in sensor-reference agreement. While the dominant particulate source remained broadly similar throughout the deployment, we agree that the current study cannot fully separate the influence of aerosol characteristics from intrinsic sensor degradation mechanisms such as optical fouling or airflow obstruction.
However, despite the known limitations of low-cost optical PM sensors in coarse-particle environments, the low-cost sensors still exhibited systematic temporal relationships with the co-located reference instrument that were sufficient for calibration modelling. We agree that the manuscript should more carefully distinguish between intrinsic sensor degradation and temporal instability in PM₁₀-equivalent sensor response relative to the reference instrument.
We also agree that the temporal structure of the calibration measurements should be more clearly described. Reference measurements in the swine deployment were collected during periodic co-location events rather than continuously throughout the study period. Accordingly, we will revise the manuscript to clarify that the estimated drift terms reflect temporal changes in sensor-reference calibration relationships across repeated sampling periods, rather than continuously observed degradation trajectories. We will additionally revise the discussion to better contextualize the limitations associated with sparse longitudinal reference measurements and annualized drift extrapolation.
Regarding PM₂.₅, co-located PM₂.₅ reference measurements were unfortunately not available during the swine deployment, limiting our ability to perform equivalent analyses for PM₂.₅ in that environment. We will clarify this limitation in the revised manuscript. Finally, we thank the reviewer for identifying the incorrect title in Section 3.3, which will be corrected from “Painting Environment” to “Swine Environment.”
Citation: https://doi.org/10.5194/egusphere-2026-1248-AC2
-
AC2: 'Reply on RC2', Oliver Stroh, 22 May 2026
Data sets
Data and Code for: Quantifying calibration drift in low-cost PM10 and sound sensors across contrasting indoor environments Oliver Stroh https://doi.org/10.17632/6dyv3bvysf.1
Model code and software
Data and Code for: Quantifying calibration drift in low-cost PM10 and sound sensors across contrasting indoor environments Oliver Stroh https://doi.org/10.17632/6dyv3bvysf.1
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 573 | 282 | 63 | 918 | 52 | 59 |
- HTML: 573
- PDF: 282
- XML: 63
- Total: 918
- BibTeX: 52
- EndNote: 59
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The submitted paper has a central theme around the measurement of sound, for which a direct link to atmospheric science, as per the definition of atmospheric science that I think is pertinent to Atmospheric Measurement Techniques, is not made. I therefore suggest the manuscript is rejected as it contains a significant volume of information that is beyond the scope of the journal. I paste relevant information from the journal's Aim and Scope description below.
It might be argued that sound can be a property of Earth's atmosphere. Even if that were accepted, I think the presented study around sound sensors is so far beyond the range of previously published articles in AMT that the current manuscript should be rejected. Suitable journals for the sound section of work are available, such as the Journal of Sound and Vibration.
Copied from https://www.atmospheric-measurement-techniques.net/about/aims_and_scope.html:
Atmospheric Measurement Techniques (AMT) is a not-for-profit international scientific journal dedicated to the publication and discussion of advances in remote sensing, as well as in situ and laboratory measurement techniques for the constituents and properties of the Earth's atmosphere.
The main subject areas comprise the development, intercomparison, and validation of measurement instruments and techniques of data processing and information retrieval for gases, aerosols, and clouds. Papers submitted to AMT must contain atmospheric measurements, laboratory measurements relevant for atmospheric science, and/or theoretical calculations of measurements simulations with detailed error analysis including instrument simulations. The manuscript types considered for peer-reviewed publication are research articles, review articles, and commentaries.
Simon P. O'Meara