All Collections
AirVisual Pro
Comparison of PM2.5 Measurements Using the Air Visual (AV) Sensor and Beta Attenuation Monitor (BAM)
Comparison of PM2.5 Measurements Using the Air Visual (AV) Sensor and Beta Attenuation Monitor (BAM)
Minhee Song avatar
Written by Minhee Song
Updated over a week ago


Measurements collected using a continuous light scattering sensor, AV, were compared to the measurements made using reference BAM data, computed by the US embassy and the Chinese government, between the 1st of June to the 30th of June 2015. The aim of this investigation was to analyze the accuracy and precision of the AV sensor in measuring the mass concentration of airborne particles with an aerodynamic diameter of less than 2.5 micrometer(µm). The PM2.5 measurements made by the AV sensor and the BAM sensor were found to correlate well, with a daily and hourly correlation efficiency of 0.96 and 0.83 respectively. As such, AV and BAM are correct for deployment in the real time continuous monitoring of PM 2.5.

Sampling Method:

All the instruments were operated in ChaoYang district, Beijing. The data from BAM was measured at the United States embassy, which is located at 55 Anjialou Road. The distance between the AV to the US embassy sensor is approximately 0.5km. The AV detector was located at a 20m-height balcony, facing a residential area, to ensure the data collected by the sensor was not affected by automotive traffic pollution.

Monitoring was conducted from the 1st of June (18:00) to the 30th of June 2015 (15:00), for a total measurement period of 30 days. The concentration of PM2.5 was monitored during June due to the high variation of the PM2.5 concentration and variety in humidity levels.

Results and discussion:

Time series plot of PM2.5 concentration

A quantitative method to compare data was used to gain a visual sense of the data’s precision by using a time series plot, as displayed in Figure 1 and 2. Average PM2.5 concentration was calculated from the data recorded every one hour. PM2.5 concentration recorded between both instruments was slightly different. This is due to different detection methods between both instruments, as AV used light scattering and BAM utilized beta ray scattering. In addition, the response time of AV and BAM is different, in which, AV was set to record the concentration every one second, while BAM records the data hourly.

A total of 694 data points were captured. While the average PM2.5 mass concentration was distributed between 0 and 250µg/m3, a similar concentration trend for AV and BAM is shown.

Figure 1: Time series of hourly average PM2.5 using AV and BAM (US Embassy)

Figure 2: Time series of hourly average PM2.5 using AV and BAM (Chinese government, agricultural exhibition center)


Two statistical analyses were used to compare concentrations from AV and BAM. One was to compare the hourly difference between the two instruments as both an absolute concentration and percentage and two was to compare AV to BAM using a linear regression analysis.

Relative difference

The relative difference can be calculated by dividing the absolute difference and the concentration value from the US embassy. By comparing the average daily measurement from the US embassy and AV sensor, the percentage difference was found to be 13.9%, which indicates an excellent correlation as the average error of the light scattering device is estimated to be approximately 30% to 40% according to (Molenar, n.d.). The relative differences are due to the natural variability of PM2.5 aerosol parameters and the scattering efficiency of the AV sensor. The spatial difference is an added contribution to the difference between the measurements. (Refer to factors that affects PM2.5 measurements)

While the hourly relative difference between AV and BAM is found to be 16.1%. Higher accuracy can be driven if you neglect outliers, especially in very low concentration levels (<8 µg/m3). The lower the concentration level, the higher the uncertainty.

Linear regression

The purpose of linear regression analysis is to explore the relationship between corresponding measurements of AV and BAM across a range of concentrations. The regression procedure determines the “best” available straight line for describing the relationship and the regression coefficient explains the correlation of the data. Figure X shows a comparison of the regression diagram.

The average coefficient of correlation squared (r2) of average daily measurements between AV and BAM was found to be 0.959. The slope was 0.9067 and the average interception was 4.6644. The agreement between the daily measurements is very good as the slope is close to 1 and r2 exceeds 0.9.

However, based on the hourly measurements, the slope is approximately 0.822 and r2 is 0.83, according to figure 4. Although the hourly data deviates more from the one on one ratiocompared to the daily measurements, r2 of 0.83 indicates a strong correlation between concentrations of AV and BAM.

While the correlation between data from the Chinese government and AV shows a high r2, with a value of 0.93, the data deviates more from a one on one ratio line more with a lower slope (0.83) and a higher intercept (9.54).

Figure 3. Linear regression lines of the average daily US embassy BAM data versus AV data

Figure 4. Linear regression lines of the average hourly US embassy BAM data vs. AV data

Figure 5. Linear regression lines of the average daily Chinese government data vs. AV data

Measurement methods

Correlation coefficient square (r2)





AV (daily)

BAM (daily)




AV (hourly)

BAM (hourly)




AV (daily)

Chinese government (daily)




Table 1. Comparison of the particle light scattering and beta ray scattering


This research and figures illustrate that the calibrated light scattering detection device, the Air Visual sensor, is useful as an alternative instrument for monitoring PM2.5 concentration levels. Despite the relative low cost of the Air Visual sensor, the results from the present study suggest that the Air Visual measurements are reasonably precise, with (R2=0.959) compared to BAM, though the Air Visual sensor may still be influenced by factors such as changes in particle characteristics.


John V. Molenar. Theoretical Analysis of PM2.5 Mass Measurements by. Nephelometry. Air Resource Specialists, Inc.

Did this answer your question?