Enhancing the Temporal Resolution and Sensitivity of Beta Attenuation Monitors via Concentrated Particle Deposition and Multi-Sensor Ensemble Averaging

Lin, Chih-Wei; Chen, Chih-Chieh; Li, Qi-Xiang; Huang, Sheng-Hsiu

doi:10.5194/egusphere-2026-3517

Preprints

https://doi.org/10.5194/egusphere-2026-3517

Preprints

29 Jun 2026

| 29 Jun 2026

Enhancing the Temporal Resolution and Sensitivity of Beta Attenuation Monitors via Concentrated Particle Deposition and Multi-Sensor Ensemble Averaging

Chih-Wei Lin, Chih-Chieh Chen, Qi-Xiang Li, and Sheng-Hsiu Huang

Abstract. Beta attenuation monitors (BAMs) are widely used for continuous ambient particulate matter (PM) monitoring, yet their effectiveness for real-time, ultra-low-concentration detection is often limited by long response times and high detection limits. The variability of the beta source intensity is the main factor for the limitation of beta gauges. To enhance BAM performance under low-concentration conditions, two complementary approaches were investigated. A commercial BAM was modified to operate with different particle deposition areas (2.0, 1.0, 0.6, and 0.4 cm²), thereby increasing the areal particle loading on the filter tape, while a multi-sensor array was simulated through ensemble averaging of repeated measurements to reduce beta intensity variability. Different temporal smoothing windows were also evaluated to examine the trade-off between measurement stability and response time. All experiments were conducted in a controlled aerosol chamber to ensure stable testing conditions. The research evaluated the effects of particle deposition area, array size, and smoothing window on response time, signal stability, and limits of detection (LOD). Experimental results indicate that reducing the particle deposition area to 0.4 cm² effectively amplifies the mass change signal per unit area, thereby increasing measurement sensitivity for active monitoring. However, this reduction also shields the detector, lowering initial beta intensity and increasing relative statistical noise. Increasing the number of measurement units to six significantly improved measurement stability, reducing the coefficient of variation (CV) of the intensity from 1.02 % to 0.39 %. A cost-benefit analysis further indicated that the marginal improvements became negligible beyond six units, suggesting an optimal array size of 4–6 measurement units. The study demonstrates that the most effective performance is achieved through a synergistic integration of these approaches. The enhanced signal provided by the reduced particle deposition area, together with the improved counting stability of the six-unit ensemble, enabled the use of a shorter 30-minute smoothing window without sacrificing measurement precision. This integrated configuration reduced the instrumental response time from 60 to 22 minutes while maintaining a standard deviation of 2.67 μg m^-3 and a detection limit of 7.75 μg m^-3. These findings provide a technical foundation for the development of high-resolution, real-time array-type beta gauges capable of meeting increasingly stringent global air quality monitoring requirements.

Received: 16 Jun 2026 – Discussion started: 29 Jun 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Chih-Wei Lin, Chih-Chieh Chen, Qi-Xiang Li, and Sheng-Hsiu Huang

Status: final response (author comments only)

CC1:
'Comment on egusphere-2026-3517', Nima Zafarmomen, 04 Jul 2026
The manuscript presents a useful technical study on improving the temporal resolution and sensitivity of beta attenuation monitors (BAMs) for real-time particulate matter monitoring. The authors investigate two main strategies: reducing the particle deposition area to increase mass loading per unit area, and using multi-sensor ensemble averaging to reduce stochastic beta-intensity variability. The results show that reducing the deposition area improves signal sensitivity, while ensemble averaging substantially improves measurement stability. The integrated configuration, using a 0.4 cm² deposition area, six-sample ensemble averaging, and a 30-minute smoothing window, reduced response time from 60 to 22 minutes while maintaining acceptable measurement precision and detection limits. Overall, the study is technically relevant and contributes to the development of higher-resolution PM monitoring systems.
The manuscript should more clearly explain the practical novelty of the proposed approach compared with existing commercial BAM designs and previous optimization studies. The authors should clarify whether the main contribution is the reduced deposition area, the multi-unit ensemble concept, or the combined optimization framework.

The experimental design is generally clear, but the simulation of a multi-sensor array using repeated measurements from a single BAM requires further justification. The authors should discuss how well repeated measurements represent truly simultaneous operation of multiple independent BAM units, especially under variable ambient conditions.

The results show a trade-off between reduced particle deposition area, increased signal sensitivity, higher pressure drop, and increased relative statistical noise. This trade-off should be discussed more explicitly in terms of practical field deployment, filter lifetime, maintenance frequency, and instrument reliability.

Several figures are important for supporting the conclusions, particularly Figures 4, 8, 10, and 11. The authors should improve figure readability by increasing font sizes, clarifying legends, and ensuring that all units and symbols are consistently formatted.

The manuscript would benefit from a clearer explanation of the detection-limit calculation. Since the authors compare empirical LOD values with theoretical estimates, the assumptions behind the LOD definition, blank measurements, smoothing window, and ensemble averaging should be stated more explicitly.

The language is generally understandable, but some sentences are long or repetitive. Careful language editing is recommended to improve grammar, flow, and conciseness, especially in the Results and Discussion section.

The authors are strongly recommended to cite recent work on high-spatiotemporal-resolution particle monitoring in complex urban environments. In particular, Yeganeh, B., Shakerdonyavi, A., Zafarmomen, N., and Taheri, A.: Comprehensive spatiotemporal analysis of long-term mobile monitoring for traffic-related particles in a complex urban environment, Atmospheric Pollution Research, 102870, 2025, is highly relevant. That study used a year-long mobile monitoring campaign to measure PM2.5 and black carbon at high spatial and temporal resolution in Tehran, demonstrating the importance of dense, high-resolution monitoring strategies for traffic-related particulate exposure assessment. Citing this work would strengthen the manuscript’s discussion of real-world applications and the need for improved high-resolution PM monitoring technologies.
Citation: https://doi.org/10.5194/egusphere-2026-3517-CC1
RC1:
'Comment on egusphere-2026-3517', Anonymous Referee #1, 26 Jul 2026
I have gone through the manuscript “Enhancing the Temporal Resolution and Sensitivity of Beta Attenuation Monitors via Concentrated Particle Deposition and Multi-Sensor Ensemble Averaging”.
The authors aimed to improve the performance of a Beta Attenuation Monitor for real-time monitoring of low concentrations of particulate matter (PM₂.₅). They investigated three optimization strategies: reducing the particle deposition area (from 2.0 to 0.4 cm²), simulating a six-unit BAM array through ensemble averaging, and optimizing the smoothing time. The optimized configuration (0.4 cm² deposition area, six-unit averaging, and a 30-minute smoothing window) reduced the instrument response time from 60 minutes to 22 minutes while maintaining a standard deviation of 2.67 μg m⁻³ and achieving an LOD of 7.75 μg m⁻³. These findings demonstrate that combining hardware optimization with statistical averaging enables faster, more sensitive, and more reliable real-time ambient air quality monitoring. The manuscript is well written.
Here are a few comments that need to be addressed before the manuscript is accepted for publication.
The abstract is too long. Please include only the best outcomes.

There is no justification provided for the selection of deposition area 2.0, 1.0, 0.6 and 0.4 cm². Please explain why the specific range has been considered.

The quality and presentation of the figures require improvement. In particular, the description provided for Figure 3 appears inconsistent with the graphical representation, suggesting that the explanation may have been reversed. The authors are advised to carefully verify the figure, labels, and corresponding discussion.

Every figure and table should be cited in the manuscript. There is no citation of Figure 4. The figure caption format should be same in manuscript. For example on page no 7 Figure 3 is written Fig. 3 and Figure 5 is written as Figure 5.

On Page 17, the manuscript refers to Equation (3); however, the equation itself is not presented. The authors should include the complete equation together with its appropriate citation and ensure that all mathematical expressions are correctly numbered and referenced.

Provide a comparative table for proposed method with other recently reported BAM optimization techniques.
Citation: https://doi.org/10.5194/egusphere-2026-3517-RC1
RC2:
'Comment on egusphere-2026-3517', Anonymous Referee #2, 27 Jul 2026
The authors seek to improve the low-concentration performance of a commercial beta attenuation monitor (Thermo 5014i) through two approaches: (1) reducing the particle deposition area (2.0 to 0.4 cm²) to increase mass loading density, and (2) simulating a multi-unit sensor array by partitioning a single instrument's continuous data record into sequential segments and ensemble-averaging those segments. The study also evaluates moving-average smoothing windows to further reduce measurement variability. The authors report reducing the response time from 60 to 22 min while achieving a reported detection limit of 7.75 µg m⁻³.
The central premise of the manuscript is that averaging sequential measurements from a single beta source and detector is representative of averaging measurements from multiple independent source-detector units. This assumption is fundamental to the study's conclusions and requires quantitative justification. Specifically, the authors should explain why sequential segments acquired from the same instrument constitute a valid surrogate for independent physical units and provide evidence that non-Poisson sources of error, such as instrumental drift, detector stability, electronic noise, source variability, tape positioning, or environmental effects, are negligible over the effective integration times considered. Without such evidence, it is unclear whether the observed improvement reflects only the reduction of random counting error or can be generalized to an ensemble of independent BAM instruments.
The manuscript has fundamental conceptual and methodological problems that undermine its central claims, along with substantial internal inconsistencies in the reported data. I detail these below.
The most significant issue:
The central motivation of this manuscript is explicitly framed around ultra-low-concentration PM2.5 monitoring. The authors cite the 2021 WHO Air Quality Guideline revision, which lowered the annual PM2.5 level to 5 µg m⁻³, and argue that ambient concentrations are increasingly approaching the operational detection limits of commercial BAMs. The stated objective of the study is therefore to enhance BAM performance under these low-concentration conditions. However, the results presented do not appear to support this central motivation.
The manuscript identifies the integrated configuration of Ts + A + N (30-min smoothing, 0.4 cm² filter area, and six-unit ensemble) as the "optimal" configuration. However, this configuration achieves a detection limit (LOD) of 7.75 µg m⁻³ (Table 3, abstract, and conclusions), which remains above the 5 µg m⁻³ WHO guideline that motivates the study. Moreover, this LOD is essentially comparable to, and slightly higher than, the baseline BAM LOD of 8.10 µg m⁻³. Thus, based on the authors' own results, the recommended integrated configuration appears to improve response time but does not provide a meaningful improvement in detection capability relative to the unmodified instrument. It therefore does not resolve the low-concentration measurement challenge that forms the basis of the manuscript.
This discrepancy directly affects the manuscript's primary claims. The abstract and conclusions state that the integrated approach satisfies the requirements for ultra-low-concentration monitoring and enables measurements at levels associated with increasingly stringent global air quality standards. However, the reported LOD of 7.75 µg m⁻³ does not demonstrate this capability. Instead, the results suggest that the proposed configuration remains unable to reliably quantify PM2.5 concentrations at or below the 5 µg m⁻³ level used to justify the work.
Other major issues:
1. The core novelty of the paper rests on simulating six independent BAM units by splitting one instrument's time series into segments and averaging them. This does not validly emulate a physical array, and the paper never justifies why it should. Averaging N sequential segments from a single source reduces variance by √N only if the segments are statistically independent and identically distributed. But sequential segments from the same physical detector share the same systematic components: the same beta source, the same detector drift, the same filter-tape geometry, the same electronic baseline. A real array of six physical detectors would have six different sources with independent systematic offsets, the very thing this simulation cannot reproduce. What the authors have actually demonstrated is little more than the textbook √N reduction of Poisson counting noise by increasing total integration time. Averaging six one-hour segments is, to first order, statistically equivalent to one six-hour measurement. The paper needs to explicitly address why this is not simply a re-labelling of longer averaging. This distinction is decisive for the whole paper. If the "array" is just longer counting, then the headline conclusions about "array-type beta gauges" and "4–6 units" are not actually supported by the data presented.
2. The quantitative results do not agree between text, tables, and figures. I cannot verify the conclusions when the same quantity takes different values in different places:
Abstract/text state the CV increased "from 1.02% to 1.97%" (p. 7, line ~150), but elsewhere it is "1.02% to 2.01%" (p. 11, line ~202) and Table 2 lists "1.85–2.01%." Which is correct?

Text says "105.6 cps" (line ~200) while Table 2 gives "102–105 cps." Figure 6 text says "102 to 118 cps."

Table 3 lists the integrated (Ts+A+N) detection limit as 7.75 µg m⁻³ and SD as 2.67, matching the abstract. But the individual entries in Table 3 contradict the running text. E.g., text (line ~300) states the six-unit-alone LOD was "1.04 µg m⁻³" but Table 3 shows "1.03"; text says area-reduction-alone SD improved to "2.4," Table 3 shows "2.40" (consistent), yet text (line ~296) says shortening smoothing gave response time 22 min while Table 3's "Ts" column shows RT of 22 with SD 12.95 and LOD 29.02, which is internally OK, but the baseline "Reference BAM" row lists Ts=75, N=1, A=2 with SD 5.54 and LOD 8.10, whereas Fig. 11 states single-sample LOD ranges 386.58 → 8.112 µg m⁻³. These need to be reconciled.

The integrated optimization is claimed as the paper's main achievement, yet Table 3 shows that N=6 alone gives SD 0.56 and LOD 1.03 µg m⁻³, far better than the "optimal" integrated configuration (SD 2.67, LOD 7.75). By the authors' own data, adding area reduction and shortening the smoothing window degrades the detection limit by nearly an order of magnitude. The narrative that the integrated approach is "the most effective" is directly refuted by the authors' own table. This must be addressed head-on; as written, the abstract's headline (7.75 µg m⁻³) is the paper's worst good-configuration LOD, not its best.

3. The claim (p. 17, line ~267) that six-sample averaging drives the LOD to 0.8 µg m⁻³ and Table 3's 0.89 for A+N should be treated with great caution.
These values are derived from the same instrument's baseline noise partitioned and averaged. As argued in point 1, they almost certainly reflect an artificial variance reduction that a real instrument, subject to source decay, temperature drift, tape-advance artifacts, relative humidity effects, and flow instability, would not achieve. Field BAMs do not reach sub-µg m⁻³ detection limits by longer averaging because they become dominated by systematic, not statistical, error at long integration times.

The paper computes LOD purely from short-term counting statistics (LOD = 2σ_blank) in a chamber. It provides no evidence of two- sample variance behaviour or the integration time beyond which drift dominates. Without such an analysis, the claim that averaging keeps lowering the LOD indefinitely is unsupported and physically suspect.

4. All experiments use one aerosol (NaCl), one size (CMD 1 µm), and one concentration (35 µg m⁻³). The paper's motivation is ultra-low concentration monitoring near 5 µg m⁻³, yet no experiment is performed at or near that level. The claimed applicability to sub-5 µg m⁻³ ambient conditions is therefore an extrapolation, not a demonstration. NaCl is non-volatile; one of the paper's stated motivations is reducing semi-volatile loss with shorter sampling. This benefit is asserted but never tested with a volatile or ambient aerosol. Z/A dependence of µ is mentioned (Jaklevic et al.) but the effect of using pure NaCl vs. ambient mixed aerosol on the mass attenuation coefficient is not discussed as a limitation.
5. Eq. (4)/(5) and the surrounding text repeatedly confuse mass units. CM is defined as "mg cm⁻³" (line ~110). This unit is nonsensical for an ambient concentration; it should be µg m⁻³. Figures 2–11 axis labels use "mg m⁻³" where the text and physical values clearly mean µg m⁻³ (e.g., "CM = 35 mg m⁻³" in Figs. 2, 4, 5, 8; 35 mg m⁻³ is 35,000 µg m⁻³, an absurd ambient level). These pervasive unit errors must be corrected everywhere. In addition, abstract line ~28 and Conclusions line ~327 report "standard deviation of 2.67" and "detection limit to 7.75" with units dropped in the conclusion.
6. The manuscript should clarify whether Eq. (2) refers to accumulated counts or count rate. As currently written, Eq. (2) corresponds to the Poisson uncertainty of the accumulated counts, whereas the surrounding discussion and the reported beta intensities (cps) suggest that the intended quantity is the uncertainty of the count rate. If beta intensity is expressed as count rate, the appropriate uncertainty is
σ_R = σ_C / t = √(Nt) / t = √(N/t).
Clarifying this distinction would improve the physical interpretation of Eq. (2) and ensure consistency between the equation and the accompanying discussion.
7. The authors conclude (p. 17, lines ~269–272) that reducing the deposition area does not provide a net LOD improvement because the increased signal is offset by reduced I_0 and higher relative noise. This is an important finding, but it conflicts with the abstract and title, which present “concentrated particle deposition” as a key enhancement mechanism. If area reduction does not improve detection performance and introduces additional drawbacks such as shorter filter lifetime and increased pressure drop, the authors should clarify its role and reconsider whether it should be highlighted as a primary contribution.
8. No replication statistics, confidence intervals, or uncertainty bars are reported on any of the summary metrics. Claims such as "marginal improvements became negligible beyond six units" and the cited "40%, 60%, 67%" cost-benefit reductions are presented without any statistical testing or error propagation. The "optimal 4–6 units" conclusion is essentially a visual judgment.
9. The manuscript repeatedly invokes "cost-benefit" considerations but does not present a quantitative cost analysis. No information is provided on instrument costs, source fabrication or procurement, maintenance, calibration, or operational overhead associated with multiple radioactive sources. Consequently, the claimed cost-benefit advantage is not substantiated. If the authors wish to retain this claim, they should either provide a quantitative cost model (or at least a comparative cost framework) or revise the discussion to describe the approach as a qualitative assessment rather than a cost-benefit analysis.
10. Figure 8 and text discuss up to 14 units, but the methods describe six repeated one-hour runs (and an 8-hour → 8-segment example). How were 14 independent segments obtained, and over what total duration? The provenance of the larger array sizes is unclear and must be specified.
11. Response Time (RT) is defined as time to 90% of steady state, and Eq. (6) (y = 0.854x − 3.96) is given without stating units or the range of validity; a negative RT results for small smoothing windows (x < 4.6), which is unphysical.
12. The literature review is relatively limited (approximately 10 references), with several key citations dating from 1981–1999. The manuscript would benefit from a more comprehensive review of recent developments in aerosol monitoring, including advances in low-cost sensor networks, ensemble or network-based PM monitoring strategies, and modern approaches to improving BAM performance (e.g., humidity correction, tape-related corrections, and hybrid monitoring approaches). In addition, the manuscript should better establish its novelty relative to prior studies that have investigated averaging or combining measurements from multiple instruments to improve precision and detection limits.
13. The experimental evaluation is limited to a single instrument model (Thermo Scientific 5014i). The manuscript should acknowledge that the results may not be directly generalizable to other commercially available beta attenuation monitors, which differ in source activity, detector design, sampling geometry, tape transport mechanisms, flow control, signal processing, and correction algorithms. Consequently, the applicability of the proposed approach beyond the Thermo 5014i remains uncertain and should be discussed as a limitation.
Citation: https://doi.org/10.5194/egusphere-2026-3517-RC2

Chih-Wei Lin, Chih-Chieh Chen, Qi-Xiang Li, and Sheng-Hsiu Huang

Viewed

Total article views: 83 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
50	25	8	83	5	7

HTML: 50
PDF: 25
XML: 8
Total: 83
BibTeX: 5
EndNote: 7

Views and downloads (calculated since 29 Jun 2026)

Month	HTML	PDF	XML	Total
Jul 2026	43	15	7	65
Aug 2026	7	10	1	18

Cumulative views and downloads (calculated since 29 Jun 2026)

Month	HTML	PDF	XML	Total
Jul 2026	43	15	7	65
Aug 2026	7	10	1	18

Viewed (geographical distribution)

Total article views: 72 (including HTML, PDF, and XML) Thereof 72 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 10 Aug 2026

Short summary

Conventional PM_2.5 Beta Attenuation Monitors (BAMs) suffer from long response times and high detection limits. This study introduces a novel methodology combining concentrated particle deposition with multi-sensor ensemble averaging to overcome these limitations. We successfully improved the instrumental response time from 60 to 22 minutes while maintaining a low detection limit of 7.75 µg m^-3. This significantly improves real-time air quality monitoring.


Total:	0
HTML:	0
PDF:	0
XML:	0