the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 19 Jan 2026
Deployment and Evaluation of a Low-Cost Sensor System for Atmospheric CO2 Monitoring on a Sea-Air Interface Buoy
Abstract. Direct in-situ observation of marine CO2 concentrations is, crucial for estimating air-sea CO2 fluxes, yet such observations remain scarce. Drawn on experiences from urban CO2 monitoring and buoy-based measurements, this study deployed a sea-air interface buoy platform in the northern South China Sea, near Maoming, Guangdong Province, China. This platform was equipped with three low-cost SenseAir K30 sensors to enable continuous atmospheric CO2 measurement. This paper presents the first detailed account of the methodology, encompassing hardware design, environmental corrections, land-based validation tests, offshore deployment procedures, and initial observational results. These findings thus provide valuable insights for advancing marine CO2 observations practices. To mitigate the impacts of temperature, humidity, and pressure on sensor readings – while simultaneously compensating for zero-drift – an environmental correction method was implemented. This approach significantly improved data accuracy: in land tests, the root mean square errors was reduced from 8.03 ppm to 3.64 ppm; in marine observations, the root mean square errors decreased from 24.26 ppm to 1.59 ppm. Importantly, this level of precision meets the requirements for resolving sea surface CO2 dynamics (~420–480 ppm). Observed concentrations were consistent with HYSPLIT-simulated long-range atmospheric transport, revealing the stable and homogeneous nature of the marine atmospheric boundary layer, with diurnal variations of approximately 3 ppm, and capturing localized or short-term fluctuations due to terrestrial carbon sources. These results demonstrate the effectiveness of the method, offering a low-cost, high-density solution for marine atmospheric CO2 monitoring and providing key inputs for inversely estimating ocean carbon sink.
- Preprint
(2148 KB) - Metadata XML
-
Supplement
(603 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-5588', Anonymous Referee #1, 03 Mar 2026
-
RC2: 'Comment on egusphere-2025-5588', Anonymous Referee #2, 18 Mar 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2025-5588/egusphere-2025-5588-RC2-supplement.pdf
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 212 | 112 | 18 | 342 | 36 | 23 | 16 |
- HTML: 212
- PDF: 112
- XML: 18
- Total: 342
- Supplement: 36
- BibTeX: 23
- EndNote: 16
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of EGUsphere-2025-5588
General comments:
Liu et al. tested and deployed low-cost NDIR carbon dioxide sensors. They successfully developed data processing, calibration and sensitivity correction protocols to improve instrument accuracy. This study is a critical first step towards lower cost monitoring options of marine carbon dioxide when using existing buoy platforms. Overall, the study is well-written and easy to follow. This manuscript is well suited for publication in AMT and will surely be of interest to experts and the wider readership of the journal. However, some details should be added when describing the data processing steps, see specific comments below. The calibration, cross-sensitivity corrections and data processing strategies appear to significantly improve sensor performance. However, the authors have not tracked this performance over the full lifetime of the sensors used. Re-testing would seem prudent. As this seems beyond the scope of this work, this limitation should be clearly stated in the manuscript to encourage future studies. Also, most figures could benefit from higher resolution. Specifically, the inlay text boxes seem somewhat pixelated.
Specific and technical comments:
Line 11: The reported numbers do not reflect the noise level of data at native temporal resolution, beyond the corrections mentioned here the data is also averaged. Should be mentioned here.
Line 173: Clarification is needed. Raw data is collected at 2 s resolution. You refer to a 4-sigma filter here, but do not specify how much data is included and at which resolution. 4-sigma of 30 data points (1 minute of observations) can be quite different than 4-sigma of 1800 observations (1 hour of observations).
Line 179: You report that the optimal averaging time is 3 minutes, however later (line 183) you decide to average the data to 1-minute averages without an explanation. Why did you decide to use a suboptimal averaging time? Is there meaningful variability in the CO2 Ocean sink (or source) below 3 minutes? Inverse modelling studies often end up using hourly averages anyways.
Line 217: Do you expect this calibration to be valid for the whole lifetime of the sensor? If not, it would seem appropriate to add a term that accounts for temporal drifts, e.g. due to the aging of the light source or decreasing sensitivity of the detector. This was included in previous studies, e.g. AMT - Characterization of a commercial lower-cost medium-precision non-dispersive infrared sensor for atmospheric CO2 monitoring in urban areas
Line 247: change - 0.39 to -0.39
Line 257: it seems some symbols got detached from the numerals
Line 263: Have measurements been repeated after several months or years to ensure the reported sensitivities to environmental parameters are consistent over time?
Line 292: there seems to be floating comma sign
Line 316: the RMSE is significantly improved and lower than on the land-based test. However, it should be discussed how much lower the variability of ambient CO2 values is in the marine test case compared to the land-based test. What is the standard deviation of the high-precision data for the testing periods?
Line 403: Are you referring to additional mixing in the surface ocean or the atmospheric boundary layer?
Line 407: Your finding suggests that the daily amplitude is 50% higher when using the CMs. How did you validate the assumption that this difference cannot cause a bias when data is used in atmospheric inverse models?
Line 445: Please state clearly if this statistic applies to the native 2 seconds resolution or for 1-minute averages.
Line 447: Please provide a reference for the concentration range of 420-480ppm. Are you referring to daily, monthly or seasonal variations here.
Line 472: What kind of long-term sensor drifts do you expect here? Aging of light source and sensor? If so, please consider adding or at least discussion the use of a coefficient for long-term drift (see comment: line 217)