Contrast Between Automated and Manual Measurements of Atmospheric PM2.5: Influences of Environmental Factors and the Improving Correction Method
1
Chengdu Ecological Environmental Monitoring Central Station of Sichuan Province, Chengdu 610066, China
2
Rural Environment Protection Engineering & Technology Center of Sichuan Province, College of Environmental Sciences, Sichuan Agricultural University, Chengdu 611130, China
*
Author to whom correspondence should be addressed.
Atmosphere 2025, 16(9), 1112; https://doi.org/10.3390/atmos16091112
Submission received: 6 August 2025
/
Revised: 16 September 2025
/
Accepted: 17 September 2025
/
Published: 22 September 2025
(This article belongs to the Section Atmospheric Techniques, Instruments, and Modeling)
Abstract
In this work, we tested the performance of automated atmospheric PM2.5 monitoring instruments and contrasted the data from automated measurements with those from filter-based reference measurements. The tested instruments include four brands of beta attenuation instruments (two were made in China, D1 and D2; the other two were imported from other countries, I1 and I2) and one brand of a light scattering instrument (also imported from another country, I3). The automated monitoring data were corrected based on the reference tests. The total testing period lasted 18 months. The objective of this work is to evaluate the influences of environmental factors on the performance of different automated instruments, and to improve the accuracy of the automated instruments by using a correction method. The results showed that contrasted with the reference tests, the absolute errors (MAE, mean absolute error; SD, standard deviation; and RMSE, root mean square error) of the automated monitoring instruments werehigher for temperature (T ≤ 10 °C), humidity (60% ≤ RH < 80%), and PM2.5 concentrations (PM2.5 ≥ 75 μg/m3). Meanwhile, the relative errors (CV, coefficient of variation; and NRMSE, normalized root mean square error) of the automated monitoring instruments were higher for humidity (RH > 80%) and PM2.5 concentrations (PM2.5 < 15 μg/m3). For winter data, it proved challenging to pass the reference test, which was based on a linear regression between 24-h average automated monitoring data and the integrated filter-based PM2.5 data (aka the KBR test). Before corrections, the pass rates of D1, D2, I1, I2, and I3 in the rolling KBR tests are 57.7%, 51.3%, 41.1%, 21%, and 90.2%, respectively. After corrections, the rates increase to 79.6%, 86.6%, 81.8%, 58.9%, and 91.8%, respectively. The coefficient corrections (corrections of system errors) have made the most prominent contribution to improving the pass rates of the winter samples. The quarterly correction method can significantly improve the data accuracy of automated monitoring instruments.
1. Introduction
PM2.5 (fine particulates with aerodynamic diameters of ≤2.5 μm) is one of key atmospheric pollutants, posing a significant threat to human health and the ecology [1,2]. PM2.5 exposure is closely related to a wide range of mental and physical health problems [2,3]. It is an urgent issue to implement continuous and efficient PM2.5 monitoring, guaranteeing real-time monitoring of air quality and safeguarding public health [4].
Automated atmospheric monitoring stations have been widely deployed in the past few decades in China [5]. Chengdu, the capital of Sichuan Province and a large city in the southwestern part of China, has established a comprehensive atmospheric monitoring network. Currently, it is convenient to compare and contrast PM2.5 monitoring data from various parts of Chengdu. The monitoring network is of great significance in air quality management and environmental protection decision-making [4,5].
Automated PM2.5 monitoring instruments have been widely applied in environmental monitoring, which are convenient and efficient but have a lower data precision and accuracy than those of the reference method [6]. The reference method is recommended by the United States Environmental Protection Agency (USEPA), using sampling filters under the actions of wind powers to capture atmospheric PM2.5 for a period of time (usually 24 h), and applying a balance to weigh the particulate matter. The reference method includes manual operation (i.e., transferring the sampling filters from the samplers to the high-precision weighing devices), which is also termed as the “manual integrated filter-based method” and “Federal Reference Method (FRM)” [7].
Beta attenuation and light scattering are two common methods used in PM2.5 automated monitoring [7]. During beta attenuation measurement, a filter strip is used to trap particulates. The beta attenuation instruments measure the concentrations of particulates by testing the loss of beta attenuation passing through the filter strips. Nowadays, beta attenuation is the most widely used automated monitoring method for PM2.5 in China and also one of the nationally recognized standard methods [8,9]. Light scattering instruments record the concentrations of PM2.5 by measuring the scattered light intensities of the particles, which have been adopted in PM2.5 measurement in recent years. Both beta attenuation and light scattering instruments are easily affected by environmental factors, such as temperature, humidity, and PM2.5 mass loading [10,11,12,13].
Chengdu features abundant rainfall, rainy days with high heat in summer, andfour distinct seasons. The error drifts of automated monitoring data influenced by climate in different seasons in Chengdu should be evaluated. Today, due to the different levels of economic development in various parts in Chengdu and the constraints of economic and social factors, the monitoring system is facing problems such as the emphasis on quantity over quality in station construction and data bias caused by multiple brands of samplers. Multiple brands of automated PM2.5 monitoring instruments have been used at different stations, coupled with inadequate management, resulting in significant data deviations.
To address the significant differences in data among different sampling stations and among different brands of samplers, and to improve the quality of PM2.5 monitoring data, we conducted a long-term contrastive study of automated and manual samplings of atmospheric PM2.5 in Chengdu. Five common brands of automated PM2.5 monitoring instruments, including four brands of beta attenuation instruments (two domestic, produced in China, D1 and D2; two imported, produced in countries other than China, I1 and I2) and one brand of a light scattering instrument (imported, I3) were run simultaneously and contrasted with manual monitoring. The performance of the automated monitoring instruments was evaluated and the key factors affecting the stability, errors, precision, and accuracy of the monitoring data were identified. A correlation method is proposed to improve the quality of automated monitoring data. This work provides reliable technical support for the automated monitoring of atmospheric PM2.5.
2. Experimental
2.1. Overall Experimental Design
The contrastive testing site is located on the rooftop of the sixth floor of the Chengdu Ecological Environment Intelligence Center, in Junping Street, in the center of Chengdu. A synchronous contrast between manual and automated monitoring was conducted on PM2.5 data for five brands instruments (D1. D2, I1, I2, and I3). There are three samplers for each brand. In total, 15 samplers were contrasted with manual measurement. The flow rates of the instruments of the five brands were all 16.7 L/min. Samplings were performed, and the data were recorded at regular intervals every day. The testing period for D1, D2, I1, I2, and manual measurements was from June 2022 to November 2023. A total of 398 sets of data were each collected. For I3, there were only 322 sets of data (starting from October 2022). The hourly meteorological parameters were obtained from the meteorological station (LUFFT, WS-600, Stuttgart, Germany) installed near the samplers.
2.2. Integrated Filter-Based PM2.5 Measurements (Reference Tests)
Three samplers (of the same brand of LVS, Comde Derenda GmbH, Wuxi, China) were used to simultaneously collect PM2.5, and three parallel samples were collected daily. The flow rates of the instruments were 16.7 L/min. Teflon filters (Φ 46.2 mm, pore size 2 μm, Cobetter, Hangzhou, China) were used for collection, starting at 10:00 a.m. and ending at 9:00 a.m. the next day. The sampling time for each sample was 23 h. If the difference in PM2.5 concentrations between parallel samples was greater than ±2%, the samples were invalid.
Before and after the samplings, the Teflon filters were maintained in a constant temperature and humidity system (CR-4, Comde Derenda GmbH, Wuxi, China) at a temperature of (20.0 °C ± 1.0 °C) and a humidity of (50 ± 5%) for 24 h. Afterwards, the filters were weighed using a millionth balance (AWS-1, Comde Derenda GmbH, Wuxi, China). After the first weighing, the second weighing was performed with an interval of 1 h. If the deviation between both weighing data was ≤0.1 mg, the weighing data was qualified. The average of both weighing data was regarded as the PM2.5 concentration of the day. Otherwise, a third weighing was performed. If the deviation between two of the three weighing data was ≤0.1 mg, the weighing was qualified. If the difference between the two data was >0.1 mg, the weighing was invalid. Moreover, two fixed standard filters were also weighed. If the deviation between the weighing data and historical data was >0.04 mg, the weighing was invalid.
The data quality standard for FRM samples set by USEPA is that the absolute and relative errors between parallel samples should be within ±2 μg/m3 and ±5% [6]. The quality control measures adopted in this work are equivalent to the standard set by EPA.
2.3. Automated PM2.5 Measurements
The Chengdu Automated Monitoring Network involves 392 automated monitoring stations. Among them, 361 used a beta attenuation method for PM2.5 measurement. We tested the performance of beta attenuation instruments (D1, D2, I1, and I2) and a light scattering instrument (I3) in a high humidity environment in Chengdu. The five brands of samplers are common in the market and have been used at different monitoring stations in Chengdu.
2.4. Performance Parameters of the Automated Instruments
The performance of the automated monitoring instruments was evaluated in this work. Seven parameters—the relative error (Er), mean bias error (MBE), mean absolute error (MAE), standard deviation (SD), coefficient of variation (CV), root mean square error (RMSE) and normalized root mean square error (NRMSE)—were calculated based on the PM2.5 mass data obtained from automated and reference monitoring. The definition equations of the seven parameters are listed in Supplementary Materials.
2.5. Influences of Environmental Factors on the Performance of Automated Instruments
Contrastive analysis was conducted on the whole set of data—the data under different ranges of temperature and humidity, the data under different PM2.5 concentrations, and the data of different seasons—to explore the impacts of environmental factors on the performance of automated instruments.
2.5.1. Performance of Automated Samplers Under Different Temperature
The obtained PM2.5 data (automated and reference) are divided into four groups (i.e., T < 10 °C, 10 °C ≤ T< 20 °C, 20 °C ≤ T < 30 °C, and T ≥ 30 °C) for contrastive analysis.
2.5.2. Performance of Automated Samplers Under Different Humidity
The obtained PM2.5 data (automated and reference) were divided into five groups (i.e., RH < 50%, 50% ≤ RH < 60%, 60% ≤ RH < 70%, 70% ≤ RH < 80% and RH ≥ 80%) for contrastive analysis.
2.5.3. Performance of Automated Samplers Under Different PM2.5 Concentration Ranges
The obtained PM2.5 data (automated and reference) were divided into four groups (i.e., PM2.5 < 15 μg/m3, 15 μg/m3 ≤ PM2.5 < 35 μg/m3, 35 μg/m3 ≤ PM2.5 < 75 μg/m3, and PM2.5 ≥ 75 μg/m3) for contrastive analysis.
2.5.4. Performance of Automated Samplers in Different Seasons
The obtained PM2.5 data (automated and reference) were divided into four groups (i.e., summer of June to August, autumn of September to November, winter of December to February in the next year, and spring of March to May) for contrastive analysis.
2.6. Method Introduction of KBR Reference Tests
A simple linear regression model can be established based on the relationship between the paired 24h averaged automated monitoring data and integrated filter-based PM2.5 data. Using the integrated filter-based PM2.5 data as the independent variable (x), and the PM2.5 data obtained from automated monitoring as the dependent variable (y), the slopes (k), intercepts (b), and coefficients (r) can be calculated based on the linear regressions [6,8]. The method is called a reference test, KBR reference test, or KBR test.
To evaluate the reliability of data from the automated instruments, the Ministry of Ecology and Environment (MEE) of China and the USEPA have developed corresponding standards, with manual integrated filter-based measurement as the references. For the automated measurements, the requirements for precision, accuracy, and the KBR tests are shown in Table 1.
According to the EPA standard [6], PM2.5 should be measured for at least 23 days per season at a location. The rolling KBR testing selects 23 consecutive days of data as a set (i.e., Day1 to Day23, Day2 to Day24, and so on and so forth) to determine whether KBR fitting between the manual and automated data meets the criteria, and to obtain the pass rates of each instrument. The pass rates of the rolling KBR tests reflect the deviation between the automated monitoring data and reference data. The higher the pass rates of the rolling KBR tests, the higher the degree of agreement between the automated monitoring data and reference data.
3. Results and Discussion
3.1. Data Efficiency
3.1.1. Efficiency of Manual Data
The samplings and weighings are performed for five workdays every week. The manual PM2.5 concentration data was evenly distributed among the different seasons. A total of 398 manual monitoring data are obtained and used for references. The efficiency of manual data is 92.71%, and the ratio of valid data to all days is 72.9%.
3.1.2. Efficiency of Automated Data
Table 2 shows the statistical results of data efficiency, operation/maintenance (O/M) vacancy rates, and instrument failure rates for the different brands of automated instruments during the testing period. O/M vacancies result from incorrect human operation and maintenance, and instrument failures result from instrument malfunction. Data efficiency is equal to 100% minus the O/M vacancy rate and the instrument failure rate.
From Table 2, the instruments of the five brands are all of lower failure rates of less than 10%, meeting the criterion given by the MEE. The instrument failure rate of I3 is the lowest (1.24%). The beta attenuation instruments with their instrument failure rates from low to high are D2 (1.51%), I2 (3.27%), I1 (8.54%) and D1(9.05%). The highest instrument failure rate is that of D1, indicating that it might be easily affected by environmental factors (i.e., PM2.5 concentrations, temperature, and humidity) during operation. Compared with the other automated instruments, D1 has poor reliability and is not suitable for long-term automated monitoring.
3.1.3. Instrument Failure Rates of Automated Instruments in Different Seasons
Table 3 shows the instrument failure rates of automated instruments in different seasons. From Table 3, the instrument failure rates are the highest in summer, with an average failure rate of 6.6%. The increase in instrument failure rates in summer might be due to frequent rainfall and high humidity. When conducting automated samplings in summer, it is necessary to pay more attention to the status of the instruments and promptly troubleshoot any faults.
3.2. PM2.5 Concentrations and Meteorological Parameters
Table 4 presents the average concentrations of PM2.5 measured by the reference method and the automated method, as well as the meteorological parameters, during the testing period. It shows that the average concentrations of PM2.5 are the highest in winter, followed by autumn, spring, and summer. The automated PM2.5 monitoring data in winter are usually lower than the manual reference data. Moreover, the annual average temperature and relative humidity are 20.9 °C and 67.3%, respectively.
3.3. Errors, Precision and Accuracy of Automated Instruments
3.3.1. Overall Performance of Automated Instruments
Seven parameters were calculated based on the PM2.5 mass data obtained from automated and reference data. The error parameters include Er, MBE, and MAE; the precision parameters include SD and RMSE; and the accuracy parameters include CV and NRMSE.
Table 5 presents the values of the seven parameters of the automated instruments. It shows that the average values of Er and MBE are negative for I1, I2, and I3, while positive for D2. Meanwhile, the average values of Er and MBE for D1 are close to zero. In this work, the imported instruments often have significant negative errors, indicating that they are less suitable for the high PM2.5 concentrations in China.
Moreover, the average SD values for the automated instruments of the five brands are all less than 5 μg/m3; the average RMSE values for the automated instruments of the five brands are all less than 7 μg/m3; the average CV values for the automated instruments of the five brands are all less than 15%; and the average NRMSE values for the automated instruments of the four brands (D1, D2, I2, I3) are less than 15%. Except for I1, the automated instruments meet the requirements given in the standards [6,8]. Moreover, the average values of the MAE, SD, CV, RMSE, and NRMSE for I1 and I2 are slightly higher than those for D1, D2, and I3.
3.3.2. Performance of Automated Instruments Under Different Temperatures
Figure 1 presents the Er, MBE, MAE, SD, CV, RMSE, and NRMSE for the automated instruments of the five brands in different temperatures. The variation trends of the seven parameters can be classified into three clusters. The first cluster includes the Er and MBE. From Figure 1a,b, it could be seen that the values of the Er and MBE increase with temperature and change gradually from negative to positive. When the temperature is lower than 10 °C, the PM2.5 concentrations in the reference tests are usually higher than those in the automated measurements, consistent with the previous studies [15,16]. The reason for this is the absorption of organic/inorganic gases and water vapor by the particles during long-term samplings [15]. FRM PM2.5 filters are conditioned at a temperature of (20.0 °C ± 1.0 °C) and humidity of (50% ± 5%) for 24 h and then weighed by balances. At the same time, each beta attenuation instrument is equipped with a smart heater which reduces the ambient RH of incoming air to 35% for removing the aerosol water content [15]. Therefore, there is residual water in the FRM samples. The particles on the filters absorb gas components and moisture during long-term sampling processes, resulting in higher PM2.5 concentrations in reference tests and obvious negative errors for the automated monitoring data contrasted with reference data.
At the temperature of (T ≥ 30 °C), the PM2.5 data obtained by the reference tests are usually lower than those in automated measurements, also consistent with the previous studies [15,16]. This is explained by the “evaporation loss” of semi-volatile components (i.e., nitrate ions) in PM2.5 during long-term sampling processes [15,16], resulting in lower PM2.5 concentrations in the reference tests and positive errors for automated monitoring data contrasted with the reference data.
The second cluster of parameters, including the MAE, SD, and RMSE, slowly decrease with temperature and then gradually stabilize. The facts indicate that the values of the errors caused by the absorption of particles during long-term sampling processes under lower temperatures are greater than those caused by the evaporator loss of particles at higher temperatures.
The third cluster of parameters, including CV and NRMSE, basically remain unchanged, suggesting that the precision and accuracy of the automated monitoring instruments are almost unchanged in different temperature ranges.
3.3.3. Performance of Automated Instruments Under Different Humidities
Figure 2 presents the Er, MBE, MAE, SD, CV, RMSE, and NRMSE for the automated instruments of the five brands at different humidities. The variation trends of the seven parameters can also be classified into three clusters. The first cluster includes the Er and MBE. The values of the Er and MBE decrease with humidity (see in Figure 2a,b). Obvious negative errors occur in higher humidity conditions (RH ≥ 80%). It has been well documented that the difference between the automated method and gravimetric method was insignificant under 80% RH [16,17,18]. However, at higher humidities with RH ≥ 80%, when the humidity exceeds the deliquescence point of ammonium sulfate, the main secondary aerosol component, the moisture absorbed by particles on filters in reference tests increases rapidly [16], leading to obvious negative errors between the automated and reference data.
The second cluster of parameters, including the MAE, SD, and RMSE, firstly increase with humidity when RH < 80% and then decrease. The values of the MAE, SD, and RMSE are greater in the conditions of 60% ≤ RH < 80% than in other humidity ranges. As humidity increases, the amount of moisture absorbed by the particles collected on the filters also increases, leading to an increase in the errors between the automated monitoring data and the reference data. However, when humidity is above 80%, it often rains. Rainfall has a scavenging effect on atmospheric particulates [19], decreasing the number of particulates and reducing the error values between the automated and reference measurements caused by moisture absorption.
The third cluster of parameters refers to the CV and NRMSE, which increase slightly with humidity. When humidity is above 80%, the increasing trends of the CV and NRMSE are more significant. The precision and accuracy of the instrument decrease under high humidity, which may be related to the adhesion of moisture on the surface of electronic components or test materials. Takahashi et al. [16] report that water content associated with particles collected on the filter strips in beta attenuation instruments do not evaporate completely in high humidity with RH ≥ 80–85%, leading to an overestimate of the PM2.5 data readings in the automated measurement.
3.3.4. Performance of Automated Instruments Under Different PM2.5 Concentration Ranges
Figure 3 presents the Er, MBE, MAE, SD, CV, RMSE, and NRMSE for the automated instruments of the five brands at different concentration ranges. The variation trends of the seven parameters can also be classified into three clusters. The first cluster of the parameters shows an approximate downward trend with an increase in the PM2.5 concentrations, including the Er and MBE. From Figure 3a,b, it could be seen that the values of the Er and MBE of the four brands of automated instruments (D1, D2, I2 and I3) decrease with PM2.5 concentrations, gradually transitioning from positive values to negative values. Meanwhile, the Er and MBE values of I1 also decrease when PM2.5 concentrations are more than 15 μg/m3. Contrasted with the reference method, the automated measurements show significant negative errors under high PM2.5 loadings. This is explained by the organic/inorganic gases and water vapor absorbed by particles on the filters during long-term sampling processes [15], which results in higher PM2.5 weighing values in the reference measurements and negative errors between the automated data and reference data.
The second cluster of parameters show an approximate upward trend with an increase of PM2.5 concentrations, including the MAE, SD, and RMSE. Similar to the process mentioned above, when PM2.5 concentrations increase, air pollution worsens, the amount of organic/inorganic gases and water absorbed by particles on filters increases, and the errors between the automated monitoring and reference tests increase with PM2.5 concentrations.
The third cluster of parameters first decreases and then stabilizes gradually with an increase in PM2.5 concentrations, including the CV and NRMSE. The precision and accuracy of the automated monitoring instruments are usually low when the PM2.5 concentrations are less than 15 μg/m3.
3.3.5. Performance of Automated Instruments in Different Seasons
Figure 4 presents the Er, MBE, MAE, SD, CV, RMSE, and NRMSE for the automated instruments of the five brands in different seasons. From Figure 4a,b, it could be found that significant negative errors occur in winter for the Er and MBE, consistent with the fact that higher PM2.5 loadings and lower temperatures are usually accompanied with obvious negative errors between the automated measurements and reference measurements. Moreover, the values of the MAE, SD, and RMSE are the highest in winter (see in Figure 4c,d,f), which is in line with the increase trends of the MAE, SD, and RMSE under high PM2.5 loadings. Furthermore, the CV and RMSE values of the automated instruments in summer are slightly higher than those in other seasons (see in Figure 4e,g), which is in agreement with the performance of CV and RMSE when PM2.5 concentrations are less than 15 μg/m3.
Overall, the impacts of PM2.5 concentrations on the error values between the automated measurements and reference measurements are more significant than those of temperature and humidity. The absolute errors (MAE, SD, and RMSE) are significant in winter, while the relative errors (CV and NRMSE) are significant in summer. During the sampling processes, more attention should be paid to the data quality in winter and summer.
3.4. KBR Reference Tests
3.4.1. Overall Test Results
Table 6 presents the KBR test results of the automated instruments of the five brands. It could be seen that the correlation coefficients (r) between the automated and manual monitoring data of the five brands of instruments are all more than 0.97, meeting the technical specification requirements given by the MEE of China [8] (≥0.95) and the EPA [6] (≥0.93). The slopes (k) of the linear regression equation for D1, D2, and I2 in summer are more than 1, but in winter, the slopes are less than 1. The slopes of I1 are generally less than 1 and are smaller than those of the other four brands, while the slopes of I3 are closest to 1 at 1.00 ± 0.03. The slopes of the beta attenuation method instruments are more significantly affected by seasonal changes.
It has been reported that the automated monitoring data can be converted into FRM data using empirical correlation equations [15]. In this work, we attempt to use correction equations between the automated and reference data to convert the automated monitoring data into reference data, thereby reducing errors between different brands of automated instruments. Each automated sampler will be first compared and contrasted with reference measurements at the central station for a period of one year. After the comparison and contrast, the correction formulas for all four seasons are obtained, and then the sampler is transferred to one of the stations to complete long-term monitoring tasks. All data obtained from the automated monitoring are converted into FRM data based on the quarterly KBR empirical formulas. By using this method, we hope to minimize the errors between the data of the different automated instruments as much as possible.
From Table 6, it can be observed that the main reason for the failure of the KBR tests is that the slopes (k) and intercepts (b) do not meet the standards. There are many similarities between the Chinese standard and the EPA standard in determining the pass rates of the KBR tests. In the following rolling KBR tests, only the Chinese national standard [8] is used to determine the pass rates.
3.4.2. Rolling KBR Test Results Before Corrections
3.4.3. Rolling KBR Test Results After Corrections
After correction, we recalculate the rolling KBR test result. The tested results of samplers are illustrated in Table 8. From Table 8, it can be seen that the pass rates of instruments from the different manufacturers have been improved after correction. The pass rates of D1, D2, I1, I2, and I3 have increased from 57.7%, 51.3%, 41.1%, 21%, and 90.2%, to 79.6%, 86.6%, 81.8%, 58.9%, and 91.8%, respectively. The slopes of the KBR linear fittings fluctuate greatly in spring and summer. The beta attenuation automated instruments were less likely to pass the KBR tests after the corrections for the spring and summer samples. However, the slopes in winter remained stable, and the pass rates had significantly improved after the corrections.
4. Conclusions
The current work contrasts the performance of automated PM2.5 measurement instruments of five brands, including four brands of beta attenuation instruments and one brand of a light scattering instrument. The manual filter-based measurements are used as the reference. The performance parameters, involving the Er, MBE, MAE, SD, RMSE, CV, and NRMSE, are calculated. KBR tests are performed, and the rolling KBR test results before and after corrections are contrasted and compared. The conclusion is obtained as follows:
- The order of instrument failure rate is as follows: I3 (1.24%) < D2 (1.51%) < I2 (3.27%) < I1 (8.54%) < D1 (9.05%). The instrument failure rates of the five brands of the automated instruments all meet the Chinese national standard, though D1 has the highest failure rate.
- The average values of the Er and MBE for I1, I2 and I3 are negative, while those of D2 are positive. Except for I1, the average values of the SD, CV, RMSE, and NRMSE for the automated instruments are consistent with the standards. Meanwhile, the average values of the MAE, SD, CV, RMSE, and NRMSE for I1 and I2 are slightly higher than those for D1, D2, and I3.
- Contrasted with manual references, the absolute errors (MAE, SD, and RMSE) of the automated monitoring instruments are higher at a temperature of (T ≤ 10 °C), humidity of (60% ≤ RH < 80%), and PM2.5 concentration of (PM2.5 ≥ 75 μg/m3). Meanwhile, the relative errors (CV and NRMSE) of the automated monitoring instruments are higher at a humidity (RH > 80%) and PM2.5 concentration of (PM2.5 ≤ 15 μg/m3).
- In the KBR tests, winter data were found to be difficult to pass. Before the corrections, the pass rates of D1, D2, I1, I2, and I3 were 57.7%, 51.3%, 41.1%, 21%, and 90.2%, respectively. After the corrections, the rates increased to 79.6%, 86.6%, 81.8%, 58.9%, and 91.8%, respectively. The coefficient corrections have made the most prominent contribution to improving the pass rates of the winter samples. The quarterly correction method can significantly improve the data accuracy of automated monitoring instruments.
Supplementary Materials
The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/atmos16091112/s1.
Author Contributions
Methodology, D.D.; validation, D.D. and J.L.; formal analysis, L.L.; investigation, D.D.; resources, K.X.; data curation, D.D. and J.L. writing—original draft preparation, D.D.; writing—review and editing, L.L.; project administration, K.X.; funding acquisition, K.X. All authors have read and agreed to the published version of the manuscript.
Funding
This work was supported by the 2023 Environmental Monitoring Network Operation Project of Sichuan Province (N5101012024000853).
Institutional Review Board Statement
Not applicable.
Informed Consent Statement
Not applicable.
Data Availability Statement
The original contributions presented in this study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author(s).
Conflicts of Interest
The authors declare no conflict of interest.
References
- Tao, W.; Song, M.; Weng, S.; Chen, X.; Cui, L. Assessing the impact of environmental regulation on ecological risk induced by PM2.5 pollution: Evidence from China. J. Clean. Prod. 2024, 451, 142029. [Google Scholar] [CrossRef]
- Thangavel, P.; Park, D.; Lee, Y. Recent insights into particulate matter (PM2.5)-mediated toxicity in humans: An overview. Int. J. Environ. Res. Public Health 2022, 19, 7511. [Google Scholar] [CrossRef] [PubMed]
- Liang, C.; Duan, F.; He, K.; Ma, Y. Review on recent progress in observations, source identifications and countermeasures of PM2.5. Environ. Int. 2016, 86, 150–170. [Google Scholar] [CrossRef] [PubMed]
- Onaiwu, G.E.; Ayidu, N.J. Advancements and innovations in PM2.5 monitoring: A comprehensive review of emerging technologies. FUDMA J. Sci. 2024, 8, 243–255. [Google Scholar] [CrossRef]
- Liu, G.; Dong, X.; Kong, Z.; Dong, K. Does national air quality monitoring reduce local air pollution? The case of PM2.5 for China. J. Environ. Manag. 2021, 296, 113232. [Google Scholar] [CrossRef] [PubMed]
- EPA 40 CFR PART 53; Ambient Air Monitoring Reference and Equivalent Methods. USEPA: Washington, DC, USA, 2010.
- Khan, T.R.; Emerson, Z.I.; Mentz, K.H. Evaluation of fine particulate matter (PM2.5) concentrations measured by collocated Federal Reference Method and Federal Equivalent Method monitors in the U.S. Atmosphere 2024, 15, 978. [Google Scholar] [CrossRef]
- HJ 817-2018; Technical Specifications for Operation and Quality Control of Ambient Air Quality Automated Monitoring System for Particulate Matter (PM10 and PM2.5). The China’s Ministry of Ecology and Environment (MEE): Beijing, China, 2018.
- HJ 653-2021; Specifications and Test Procedures for Ambient Air Quality Continuous Automated Monitoring System for PM10 and PM2.5. The China’s Ministry of Ecology and Environment (MEE): Beijing, China, 2021.
- Samae, H.; Suriyawong, P.; Yawootti, A.; Phairuang, W.; Sampattagul, S. Precision and accuracy analysis of PM2.5 light-scattering sensor: Field and laboratory experiments. Atmosphere 2025, 16, 76. [Google Scholar] [CrossRef]
- Kashuba, R.; Scheff, P.A. Nonlinear regression adjustments of multiple continuous monitoring methods produce effective characterization of short-term fine particulate matter. J. Air Waste Manag. Assoc. 2008, 58, 812–820. [Google Scholar] [CrossRef] [PubMed]
- Barkjohn, K.K.; Gantt, B.; Clements, A.L. Development and application of a United States-wide correction for PM2.5 data collected with the PurpleAir sensor. Atmos. Meas. Tech. 2021, 14, 4617–4637. [Google Scholar] [CrossRef] [PubMed]
- Tsai, C.J.; Chang, C.T.; Huang, C.H. Direct field observation of the relative humidity effect on the beta-gauge readings. J. Air Waste Manag. Assoc. 2006, 56, 834–840. [Google Scholar] [CrossRef] [PubMed]
- Duvall, R.M.; Clements, A.L.; Hagler, G.; Kamal, A.; Kilaru, V.; Goodman, L.; Frederick, S.; Barkjohn, K.K.; VonWald, I.; Greene, D.; et al. Performance Testing Protocols, Metrics, and Target Values for Fine Particulate Matter Air Sensors: Use in Ambient, Outdoor, Fixed Site, Non-Regulatory Supplemental and Informational Monitoring Applications; US EPA Office of Research and Development: Washington, DC, USA, 2021. [Google Scholar]
- Le, T.C.; Shukla, K.K.; Chen, Y.T.; Chang, S.C.; Lin, T.Y.; Li, Z.; Pui, D.Y.H.; Tsai, C.J. On the concentration differences between PM2.5 FEM monitors and FRM samplers. Atmos. Environ. 2020, 222, 117138. [Google Scholar] [CrossRef]
- Takahashi, K.; Minoura, H.; Sakamoto, K. Examination of discrepancies between beta-attenuation and gravimetric methods for the monitoring of particulate matter. Atmos. Environ. 2008, 42, 5232–5240. [Google Scholar] [CrossRef]
- Shukla, K.; Aggarwal, S.G. A technical overview on beta-attenuation method for the monitoring of particulate matter in ambient air. Aerosol Air Qual. Res. 2022, 22, 220195. [Google Scholar] [CrossRef]
- Shi, J.; Chen, F.; Cai, Y.; Fan, S.; Cai, J.; Chen, R.; Kan, H.; Lu, Y.; Zhao, Z. Validation of a light-scattering PM2.5 sensor monitor based on the long-term gravimetric measurements in field tests. PLoS ONE 2017, 12, e0185700. [Google Scholar] [CrossRef] [PubMed]
- Kulshrestha, U.C.; Reddy, L.A.K.; Satyanarayana, J.; Kulshrestha, M.J. Real-time wet scavenging of major chemical constituents of aerosols and role of rain intensity in Indian region. Atmos. Environ. 2009, 43, 5123–5127. [Google Scholar] [CrossRef]
Figure 1.
Performance parameters of the automated instruments in different temperatures. (a) Er; (b) MBE; (c) MAE; (d) SD; (e) CV; (f) RMSE; and (g) NRMSE. Note: Multiple comparisons were calculated using the SPSS v21.0 software. The upper-case letter indicates that the mean difference is significant at the 0.01 level (p < 0.01); and the lower-case letter indicates that at the 0.05 level (p < 0.05). Arrows indicate the overall variation trends of the parameters.
Figure 1.
Performance parameters of the automated instruments in different temperatures. (a) Er; (b) MBE; (c) MAE; (d) SD; (e) CV; (f) RMSE; and (g) NRMSE. Note: Multiple comparisons were calculated using the SPSS v21.0 software. The upper-case letter indicates that the mean difference is significant at the 0.01 level (p < 0.01); and the lower-case letter indicates that at the 0.05 level (p < 0.05). Arrows indicate the overall variation trends of the parameters.
Figure 2.
Performance parameters of the automated monitoring instruments at different humidities. (a) Er; (b) MBE;(c) MAE; (d) SD; (e) CV; (f) RMSE; and (g) NRMSE. Note: Multiple comparisons were calculated using the SPSS v21.0 software. The upper-case letter indicates that the mean difference is significant at the 0.01 level (p < 0.01); and the lower-case letter indicates that at the 0.05 level (p < 0.05). Arrows indicate the overall variation trends of the parameters.
Figure 2.
Performance parameters of the automated monitoring instruments at different humidities. (a) Er; (b) MBE;(c) MAE; (d) SD; (e) CV; (f) RMSE; and (g) NRMSE. Note: Multiple comparisons were calculated using the SPSS v21.0 software. The upper-case letter indicates that the mean difference is significant at the 0.01 level (p < 0.01); and the lower-case letter indicates that at the 0.05 level (p < 0.05). Arrows indicate the overall variation trends of the parameters.
Figure 3.
Performance parameters of automated monitoring instruments at different PM2.5 concentrations. (a) Er; (b) MBE;(c) MAE; (d) SD; (e) CV; (f) RMSE; (g) NRMSE. Arrows indicate the overall variation trends of the parameters.
Figure 3.
Performance parameters of automated monitoring instruments at different PM2.5 concentrations. (a) Er; (b) MBE;(c) MAE; (d) SD; (e) CV; (f) RMSE; (g) NRMSE. Arrows indicate the overall variation trends of the parameters.
Figure 4.
Performance parameters of the automated monitoring instruments in different seasons. (a) Er; (b) MBE; (c) MAE; (d) SD; (e) CV; (f) RMSE; (g) NRMSE.
Figure 4.
Performance parameters of the automated monitoring instruments in different seasons. (a) Er; (b) MBE; (c) MAE; (d) SD; (e) CV; (f) RMSE; (g) NRMSE.
Table 1.
The requirements for accuracy, precision, and KBR tests of automated instruments in Chinese national standard and EPA standards.
Table 1.
The requirements for accuracy, precision, and KBR tests of automated instruments in Chinese national standard and EPA standards.
| Item | USEPA Standard [6,14] | Chinese National Standard [8] | |
|---|---|---|---|
| Instrument failure rate | / | <10% | |
| Precision | CV | / | ≤15% |
| NRMSE | ≤15% [6] | / | |
| Accuracy | SD | ≤5 μg/m3 [14] | / |
| RMSE | ≤7 μg/m3 [14] | / | |
| KBR tests | Slope (k) | 1.0 ± 0.10 [6] | 1.0 ± 0.10 |
| Intercept (b) | −2 μg/m3 ≤ b ≤ 2 μg/m3 and (15.05–17.32k) μg/m3 ≤ b ≤ (15.05–13.20k) μg/m3 [6] | If k ≥ 1, −5 μg/m3 ≤ b ≤ (55–50k) μg/m3 If k < 1, (45–50k) μg/m3 ≤ b ≤ 5 μg/m3 | |
| Correlation coefficient (r) | r ≥ 0.93 [6] | r ≥ 0.95 | |
Table 2.
Number of efficient data and data efficiency of automated and manual measurements.
| Manual or Brands of Automated Samplers | Manual | D1 | D2 | I1 | I2 | I3 |
|---|---|---|---|---|---|---|
| Total number of data | 398 | 398 | 398 | 398 | 398 | 322 |
| Number of efficient data | 369 | 345 | 378 | 347 | 368 | 303 |
| Data efficiency | 92.71% | 86.68% | 94.97% | 87.19% | 92.46% | 94.10% |
| Number of O/M vacancies | 15 | 17 | 14 | 17 | 17 | 15 |
| O/M vacancy rate | 3.77% | 4.27% | 3.52% | 4.27% | 4.27% | 4.66% |
| Number of instrument failures | 14 | 36 | 6 | 34 | 13 | 4 |
| Instrument failure rate | 3.52% | 9.05% | 1.51% | 8.54% | 3.27% | 1.24% |
Table 3.
Instrument failure rates of automated instruments under different seasons.
| Season | Number of Sampling Days | D1 | D2 | I1 | I2 | I3 | Ave. Instrument Failure Rate | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| a | b | a | b | a | b | a | b | a | b | |||
| Spring | 75 | 2 | 2.7% | 0 | 0.0% | 9 | 12.0% | 0 | 0.0% | 0 | 0.0% | 2.9% |
| Summer | 131/79 c | 21 | 16.0% | 3 | 2.3% | 18 | 13.7% | 1 | 0.8% | 0 | 0.0% | 6.6% |
| Autumn | 129/105 c | 13 | 10.1% | 2 | 1.6% | 3 | 2.3% | 7 | 5.4% | 1 | 1.0% | 4.1% |
| Winter | 63 | 0 | 0.0% | 1 | 1.6% | 4 | 6.3% | 5 | 7.9% | 3 | 4.8% | 4.1% |
a: Number of the instrument failure days. b: Instrument failure rate. c: The former data is for beta attenuation instruments of D1, D2, I1, and I2, and the latter data for light scattering instrument of I3.
Table 4.
PM2.5 concentrations and meteorological parameters during the testing period.
| Item | Summer | Autumn | Winter | Spring | Annual |
|---|---|---|---|---|---|
| Temperature (°C) | 28.6 ± 3.1 | 19.3± 4.4 | 9.6 ± 2.4 | 20.2 ± 4.6 | 20.9 ± 7.4 |
| Relative humidity (%) | 65.8 ± 10.4 | 71.3 ± 9.3 | 65.8 ± 10.8 | 64.3 ± 10.5 | 67.3 ± 10.5 |
| PM2.5 concentration (manual, μg/m3) | 23.0 ± 10.2 | 40.1 ± 22.3 | 65.4 ± 27.9 | 33.3 ± 15.9 | 36.8 ± 23.5 |
| PM2.5 concentration (D1, μg/m3) | 24.9 ± 11.3 | 39.0 ± 21.0 | 57.4 ± 25.0 | 35.2 ± 16.1 | 36.9 ± 21.2 |
| PM2.5 concentration (D2, μg/m3) | 25.0 ± 11.6 | 40.6 ± 23.8 | 60.0 ± 26.4 | 35.4 ± 17.0 | 37.3 ± 22.7 |
| PM2.5 concentration (I1, μg/m3) | 20.8 ± 9.7 | 34.3 ± 20.0 | 57.3 ± 26.7 | 29.6 ± 15.7 | 32.7 ± 21.5 |
| PM2.5 concentration (I2, μg/m3) | 23.9 ± 12.1 | 37.5 ± 20.1 | 59.4 ± 26.0 | 35.5 ± 18.7 | 35.6 ± 21.6 |
| PM2.5 concentration (I3, μg/m3) | 21.1 ± 10.8 | 39.8 ± 22.9 | 63.9 ± 29.0 | 33.8 ± 16.3 | 37.9 ± 24.7 |
Table 5.
Performance parameters of the automated instruments during the testing period.
| Item | D1 | D2 | I1 | I2 | I3 |
|---|---|---|---|---|---|
| Number of data | 328 | 356 | 324 | 346 | 294 |
| Er (%) | 0.5 ± 12.8 | 3.2 ± 13.7 | −13.5 ± 11.0 | −1.8 ± 15.8 | −3.5 ± 10.3 |
| MBE (μg/m3) | −1.2 ± 5.1 | 0.4 ± 4.3 | −4.6 ± 4.0 | −1.3 ± 6.0 | −1.0 ± 2.9 |
| MAE (μg/m3) | 3.7 ± 3.7 | 3.2 ± 3.0 | 4.9 ± 3.7 | 4.4 ± 4.3 | 2.3 ± 2.0 |
| SD (μg/m3) | 1.2 ± 0.8 | 1.6 ± 1.6 | 3.1 ± 2.3 | 2.4 ± 2.2 | 1.8 ± 1.9 |
| CV (%) | 4.1 ± 3.7 | 4.9 ± 3.4 | 11.4 ± 8.8 | 7.0 ± 5.9 | 5.3 ± 3.9 |
| RMSE (μg/m3) | 4.0 ± 3.6 | 3.6 ± 3.1 | 5.7 ± 3.8 | 5.1 ± 4.3 | 3.0 ± 2.2 |
| NRMSE (%) | 11.0 ± 8.0 | 11.3 ± 9.8 | 17.5 ± 9.8 | 14.6 ± 9.9 | 9.3 ± 7.6 |
Table 6.
KBR test results for the different brands of automated samplers in different seasons.
| Brand | Sampling Time | k | b | r | Pass/Not Pass Standard ① | Reason for Not Passing the Test | Pass/Not Pass Standard ② | Reason for not Passing the Test |
|---|---|---|---|---|---|---|---|---|
| D1 | Summer, 2022 (n = 26) | 1.114 | −1.500 | 0.987 | No | k > 1.1 | No | k > 1.1 |
| Autumn, 2022 (n = 37) | 0.813 | 4.309 | 0.974 | No | k < 0.9, b < 45–50k | No | k < 0.9, b > 2 | |
| Winter, 2022 (n = 54) | 0.880 | −0.263 | 0.989 | No | k < 0.9, b < 45–50k | No | k < 0.9, b < 15.05–13.72k | |
| Spring, 2023 (n = 70) | 0.977 | 2.288 | 0.974 | Yes | No | b > 2 | ||
| Summer, 2023 (n = 75) | 1.020 | 1.350 | 0.982 | Yes | Yes | |||
| Autumn, 2023 (n = 67) | 0.937 | 1.488 | 0.992 | Yes | Yes | |||
| Total (n = 329) | 0.862 | 4.071 | 0.983 | No | k < 0.9 | No | k < 0.9, b > 2 | |
| D2 | Summer, 2022 (n = 42) | 1.185 | −0.290 | 0.983 | No | k > 1.1, b > 55–50k | No | k > 1.1 |
| Autumn, 2022 (n = 45) | 0.976 | 1.265 | 0.986 | Yes | Yes | |||
| Winter, 2022 (n = 53) | 0.940 | −2.360 | 0.993 | No | b < 45–50k | No | b < −2 | |
| Spring, 2023 (n = 74) | 1.046 | 0.344 | 0.975 | Yes | Yes | |||
| Summer, 2023 (n = 77) | 1.063 | −0.188 | 0.983 | Yes | Yes | |||
| Autumn, 2023 (n = 68) | 1.065 | −1.719 | 0.995 | Yes | Yes | |||
| Total (n = 359) | 0.940 | 2.587 | 0.983 | Yes | No | b > 2 | ||
| I1 | Summer, 2022 (n = 42) | 0.947 | −1.206 | 0.978 | Yes | Yes | ||
| Autumn, 2022 (n = 47) | 0.802 | 0.545 | 0.975 | No | k < 0.9, b < 45–50k | No | b < 15.05–13.72 k | |
| Winter, 2022 (n = 50) | 0.948 | −5.156 | 0.994 | No | b < 45–50k | No | b < −2 | |
| Spring, 2023 (n = 61) | 0.923 | −0.497 | 0.976 | Yes | Yes | |||
| Summer, 2023 (n = 61) | 0.941 | −1.079 | 0.983 | Yes | Yes | |||
| Autumn, 2023 (n = 63) | 0.935 | −1.878 | 0.996 | No | b < 45–50k | No | b < 15.05–13.72 k | |
| Total (n = 324) | 0.893 | −0.644 | 0.989 | No | k < 0.9, b < 45–50k | No | k < 0.9, b < 15.05–13.72 k | |
| I2 | Summer, 2022 (n = 43) | 1.021 | 0.041 | 0.974 | Yes | Yes | ||
| Autumn, 2022 (n = 42) | 0.809 | 2.859 | 0.987 | No | k < 0.9, b < 45–50k | No | k < 0.9, b > 2 | |
| Winter, 2022 (n = 49) | 0.873 | 1.499 | 0.979 | No | k < 0.9 | No | k < 0.9 | |
| Spring, 2023 (n = 73) | 1.134 | −2.549 | 0.967 | No | k > 1.1 | No | b < −2 | |
| Summer, 2023 (n = 75) | 1.141 | −2.186 | 0.928 | No | k > 1.1 | No | b < −2 | |
| Autumn, 2023 (n = 67) | 0.891 | 2.261 | 0.980 | No | k < 0.9 | No | b > 2 | |
| Total (n = 349) | 0.884 | 2.933 | 0.970 | No | k < 0.9 | No | b > 2 | |
| I3 | Autumn, 2022 (n = 32) | 0.942 | 2.365 | 0.993 | Yes | No | b > 2 | |
| Winter, 2022 (n = 52) | 1.019 | −2.010 | 0.992 | Yes | No | b < −2 | ||
| Spring, 2023 (n = 73) | 1.013 | −0.640 | 0.972 | Yes | Yes | |||
| Summer, 2023 (n = 74) | 1.000 | −1.759 | 0.990 | Yes | Yes | |||
| Autumn, 2023 (n = 66) | 0.976 | −0.670 | 0.997 | Yes | Yes | |||
| Total (n = 297) | 1.002 | −1.091 | 0.993 | Yes | Yes |
Table 7.
Rolling KBR pass rates of the automated instruments before corrections.
| Brand | Season | Number of the Data | k Judgment/ Passing Times (Pass Rate) | b Judgment/ Passing Times (Pass Rate) | r Judgment/ Passing Times (Pass Rate) | KBR Result/ Passing Times (Pass Rate) |
|---|---|---|---|---|---|---|
| D1 | Spring | 47 | 42 (89.4%) | 44 (93.6%) | 43 (91.5%) | 40 (85.1%) |
| Summer | 57 | 25 (43.9%) | 44 (77.2%) | 57 (100%) | 25 (43.9%) | |
| Autumn | 60 | 47 (78.3%) | 53 (88.3%) | 60 (100%) | 47 (78.3%) | |
| Winter | 32 | 3 (9.4%) | 3 (9.4%) | 32 (100%) | 1 (3.1%) | |
| All | 196 | 117 (59.7%) | 144 (73.5%) | 192 (98%) | 113 (57.7%) | |
| D2 | Spring | 49 | 35 (71.4%) | 45 (91.8%) | 45 (91.8%) | 31 (63.3%) |
| Summer | 75 | 31 (41.3%) | 47 (62.7%) | 74 (98.7%) | 30 (40%) | |
| Autumn | 69 | 46 (66.7%) | 67 (97.1%) | 69 (100%) | 46 (66.7%) | |
| Winter | 31 | 31 (100%) | 8 (25.8%) | 31 (100%) | 8 (25.8%) | |
| All | 224 | 143 (63.8%) | 167 (74.6%) | 219 (97.8%) | 115 (51.3%) | |
| I1 | Spring | 39 | 23 (59%) | 19 (48.7%) | 37 (94.9%) | 17 (43.6%) |
| Summer | 59 | 49 (83.1%) | 53 (89.8%) | 59 (100%) | 49 (83.1%) | |
| Autumn | 66 | 47 (71.2%) | 13 (19.7%) | 66 (100%) | 13 (19.7%) | |
| Winter | 28 | 28 (100%) | 0 (0%) | 28 (100%) | 0 (0%) | |
| All | 192 | 147 (76.6%) | 85 (44.3%) | 190 (99.0%) | 79 (41.1%) | |
| I2 | Spring | 48 | 5 (10.4%) | 13 (27.1%) | 45 (93.8%) | 2 (4.2%) |
| Summer | 74 | 52 (70.3%) | 51 (68.9%) | 29 (39.2%) | 29 (39.2%) | |
| Autumn | 65 | 26 (40%) | 38 (58.5%) | 53 (81.5%) | 14 (21.5%) | |
| Winter | 27 | 13 (48.1%) | 6 (22.2%) | 27 (100%) | 0 (0%) | |
| All | 214 | 96 (44.9%) | 108 (50.5%) | 154 (72%) | 45 (21%) | |
| I3 | Spring | 48 | 45 (93.8%) | 46 (95.8%) | 46 (95.8%) | 43 (89.6%) |
| Summer | 52 | 48 (92.3%) | 48 (92.3%) | 52 (100%) | 48 (92.3%) | |
| Autumn | 54 | 54 (100%) | 54 (100%) | 54 (100%) | 54 (100%) | |
| Winter | 30 | 30 (100%) | 21 (70%) | 30 (100%) | 21 (70%) | |
| All | 184 | 177 (96.2%) | 169 (91.8%) | 182 (98.9%) | 166 (90.2%) |
Table 8.
Rolling KBR pass rates of automated instruments after corrections.
| Brand | Season | Number of the Data | k Judgment/ Passing Times (Pass Rate) | b Judgment/ Passing Times (Pass Rate) | r Judgment/ Passing Times (Pass Rate) | KBR Result/ Passing Times (Pass Rate) |
|---|---|---|---|---|---|---|
| D1 | Spring | 47 | 32 (68.1%) | 47 (100%) | 43 (91.5%) | 30 (63.8%) |
| Summer | 57 | 36 (63.2%) | 51 (89.5%) | 57 (100%) | 36 (63.2%) | |
| Autumn | 60 | 58 (96.7%) | 60 (100%) | 60 (100%) | 58 (96.7%) | |
| Winter | 32 | 32 (100%) | 32 (100%) | 32 (100%) | 32 (100%) | |
| All | 196 | 158 (80.6%) | 190 (96.7%) | 192 (98%) | 156 (79.6%) | |
| D2 | Spring | 49 | 45 (91.8%) | 49 (100%) | 45 (91.8%) | 43 (87.8%) |
| Summer | 75 | 53 (70.7%) | 55 (73.3%) | 74 (98.7%) | 52 (69.3%) | |
| Autumn | 69 | 68 (98.6%) | 69 (100%) | 69 (100%) | 68 (98.6%) | |
| Winter | 31 | 31 (100%) | 31 (100%) | 31 (100%) | 31 (100%) | |
| All | 224 | 197 (87.9%) | 204 (91.1%) | 199 (97.5%) | 194 (86.6%) | |
| I1 | Spring | 39 | 31 (79.5%) | 39 (100%) | 37 (94.9%) | 29 (74.4%) |
| Summer | 59 | 43 (72.9%) | 59 (100%) | 59 (100%) | 43 (72.9%) | |
| Autumn | 70 | 60 (85.7%) | 66 (94.3%) | 68 (97.1%) | 59 (84.3%) | |
| Winter | 28 | 28 (100%) | 28 (100%) | 28 (100%) | 28 (100%) | |
| All | 196 | 162 (82.7%) | 192 (98%) | 192 (98%) | 159 (81.1%) | |
| I2 | Spring | 48 | 37 (77.1%) | 46 (95.8%) | 45 (93.8%) | 34 (70.8%) |
| Summer | 74 | 31 (41.9%) | 46 (62.2%) | 29 (39.2%) | 24 (32.4%) | |
| Autumn | 65 | 61 (93.8%) | 61 (93.8%) | 53 (81.5%) | 53 (81.5%) | |
| Winter | 27 | 27 (100%) | 15 (55.6%) | 27 (100%) | 15 (55.6%) | |
| All | 214 | 156 (72.9%) | 168 (78.5%) | 154 (72%) | 126 (58.9%) | |
| I3 | Spring | 48 | 39 (81.3%) | 46 (95.8%) | 46 (95.8%) | 37 (77.1%) |
| Summer | 52 | 48 (92.3%) | 52 (100%) | 52 (100%) | 48 (92.3%) | |
| Autumn | 54 | 54 (100%) | 54 (100%) | 54 (100%) | 54 (100%) | |
| Winter | 30 | 30 (100%) | 30 (100%) | 30 (100%) | 30 (100%) | |
| All | 184 | 171 (92.9%) | 182 (98.9%) | 182 (98.9%) | 169 (91.8%) |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
MDPI and ACS Style
Dai, D.; Li, J.; Xiao, K.; Li, L. Contrast Between Automated and Manual Measurements of Atmospheric PM2.5: Influences of Environmental Factors and the Improving Correction Method. Atmosphere 2025, 16, 1112. https://doi.org/10.3390/atmos16091112
AMA Style
Dai D, Li J, Xiao K, Li L. Contrast Between Automated and Manual Measurements of Atmospheric PM2.5: Influences of Environmental Factors and the Improving Correction Method. Atmosphere. 2025; 16(9):1112. https://doi.org/10.3390/atmos16091112
Chicago/Turabian StyleDai, Dongjue, Jingang Li, Kuang Xiao, and Li Li. 2025. "Contrast Between Automated and Manual Measurements of Atmospheric PM2.5: Influences of Environmental Factors and the Improving Correction Method" Atmosphere 16, no. 9: 1112. https://doi.org/10.3390/atmos16091112
APA StyleDai, D., Li, J., Xiao, K., & Li, L. (2025). Contrast Between Automated and Manual Measurements of Atmospheric PM2.5: Influences of Environmental Factors and the Improving Correction Method. Atmosphere, 16(9), 1112. https://doi.org/10.3390/atmos16091112
Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.
