While low-cost particle sensors are increasingly being
used in numerous applications, most of them have no heater or dryer at the
inlet to remove water from the sample before measurement. Deliquescent
growth of particles and the formation of fog droplets in the atmosphere can
lead to significant increases in particle number concentration (PNC) and
mass concentrations reported by such sensors. We carried out a detailed
study using a Plantower PMS1003 low-cost particle sensor, both in the
laboratory and under actual ambient field conditions, to investigate its
response to increasing humidity and the presence of fog in the air. We found
significant increases in particle number and mass concentrations at relative
humidity above about 75 %. During a period of fog, the total PNC increased
by 28 %, while the PNC larger than 2.5
The rapid technological advancements in the fields of material science, digital electronics, and wireless communication have given rise to a wide range of low-cost air quality sensors that are now readily available on the market. These sensors are increasingly being used in many applications that were previously not achievable with conventional expensive equipment (Kumar et al., 2015; Rai et al., 2017; Snyder et al., 2013). Some of these applications are the monitoring of personal exposure and indoor air pollution and the gathering of high-resolution spatio-temporal air pollution data by means of extensive sensor networks. The data thus derived are being utilised for a variety of air pollution management tasks such as supplementing conventional air pollution monitoring, understanding the link between pollutant exposure and human health, emergency response management, hazardous leak detection and source compliance monitoring. In the process, they also serve to increase the community's awareness and engagement towards air quality issues (Snyder et al., 2013; Jovasevic-Stojanovic et al., 2015; Rai et al., 2017).
However, there are many questions regarding the reliability and, in particular, the accuracy of these low-cost sensors and their suitability in the applications that they are being used (Lewis and Edwards, 2016). Many of these sensors have serious limitations. For example, while many particle sensors respond well to high concentrations, they fail to do so at lower levels such as typical ambient concentrations (Jayaratne et al., 2018; Rai et al., 2017; Kelly et al., 2017). Single gas sensors are very often affected by other interfering gases (Fine et al., 2010; Piedrahita et al., 2014), while environmental parameters, such as temperature and humidity, can also affect the performance of these sensors under certain conditions (Holstius et al., 2014; Rai et al., 2017; Crilley et al., 2018; Jayaratne et al., 2018).
In this paper, we investigate the effect of atmospheric relative humidity on the performance of a low-cost particulate matter sensor. Humid conditions can affect the performance of a sensor in several ways. For example, sensors that operate on the principle of light scattering are affected, as the particle refractive indices are dependent on relative humidity (Hänel, 1972; Hegg et al., 1993). High humidity can cause condensation to form on electrical components, leading to resistive bridges across components. In gas sensors, condensation on the sensor surfaces can affect the reactions that give rise to the measurable electric currents.
Hygroscopic growth occurs when the relative humidity exceeds the deliquescence point of a substance. There are many hygroscopic salts such as sodium chloride, that absorb water and grow at relative humidity as low as 70 %, present in the atmosphere, especially in marine environments (Hu et al., 2010). Jamriska et al. (2008) found a significant effect of relative humidity on traffic emission particles in the size range 150–880 nm and attributed it to hygroscopic particle growth. Crilley et al. (2018) demonstrated a significantly large positive artefact in measured particle mass by an Alphasense OPC-N2 sensor during times of high ambient relative humidity. Manikonda et al. (2016) cautioned against using PM sensors in outdoor locations at high humidity due to hygroscopic growth of particles. In circumstances where the relative humidity approaches 100 %, there is the possibility of mist or fog droplets that are detected as particles. While there is a causal link between particle pollution and adverse human health effects, the presence of water on the particles plays no part in it. Therefore, air quality standards for particles are based on the dry, solid material only, and stipulate that the liquid portion must be eliminated when measuring particle mass for regulatory purposes. In order to achieve this, many conventional particle mass monitors, such as the standard tapered element oscillating microbalance (TEOM), employ a charcoal heater at its inlet to remove all liquids from the particles that are being measured (Charron et al., 2004; Alexandrova et al., 2003). Thus, sensors with no drying facility at the inlet measure what is actually present in the environment rather than what is required under regulatory protocols.
The composition of particles in the atmosphere of Brisbane, as derived from Harrison (2007), is shown in Fig. 1. The subtropical, near-coastal environment is characterised by the presence of several hygroscopic salts such as sodium chloride, ammonium sulfate, and ammonium nitrate that have deliquescence relative humidities in the range of 70–80 % (Hu et al., 2010). Many particles in the air in Brisbane contain these salts in varying concentrations. Once the relative humidity exceeds the respective deliquescence values, those salts begin to absorb water, resulting in particle growth and the excess water registered by PM sensors, unless they are removed at the instrument inlets by heating or drying (Alexandrova et al., 2003). While more expensive instruments, such as the TEOM, have built-in drying features at the sample inlets, it is not standard on low-cost sensors and even in many other mid-cost monitors such as the TSI DustTrak (Kingham et al., 2006).
Composition of particles in the atmosphere of Brisbane, as derived from Harrison (2007).
There have been very few studies of the effect of relative humidity on the performance of low cost sensors. Wang et al. (2015) investigated the performance of three low cost particle sensors based on light scattering and concluded that the absorption of infrared radiation by a film of water on a particle can cause an overestimation of the derived particle mass concentration due to the reduced intensity of light received by the phototransistor. Hojaiji et al. (2017) showed that the particle mass concentration reported by a Sharp PM sensor increased when the humidity was increasing but not when it was decreasing. While several studies have drawn attention to a possible effect of humidity on the performance of low cost sensors, no study has reliably quantified the effect. This study was carried out to investigate and to assess the magnitude of the effect of relative humidity on the performance of a low-cost particle sensor and to understand the mechanisms involved.
In this study, we focussed on the effect of relative humidity on the performance of a low-cost particle sensor in the laboratory and under real world conditions in an outdoor location at an air quality monitoring station with standard instrumentation.
Prior to commencing this study we tested a range of low-cost particle
sensors, including the Sharp GP2Y, Shinyei PPD42NS, Plantower PMS1003,
Innociple PSM305 and the Nova SDS011 (Jayaratne et al., 2018). All of them
were found to be affected to some degree by humidity with the Sharp and
Shinyei being affected at relative humidity as low as 50 % while the other
three showed deviations from the standard instruments when the relative
humidity exceeded 75–80 %. Considering their performance characteristics,
the Plantower PMS1003 was selected as the most suitable sensor for this
study. This sensor was selected because it is freely available, low-cost
(around AUD 20) and its performance characteristics have been previously
investigated extensively in our laboratories and found to be superior to the
other sensors tested (Jayaratne et al, 2018). The PMS1003 is a compact
particle sensor that monitors particles larger than 0.3
The PMS1003 was mounted on a custom interface board including a low-power microcontroller with multiple serial interfaces, a high-resolution 16-bit analog to digital converter, and a real-time clock that provided accurate time-stamping of the measurements. The PMS1003 was attached to a frame along with the interface board, allowing unobstructed airflow into and out of the device. The microcontroller was programmed to perform the necessary signal processing and power management. The time-stamped data were transferred in real-time via USB serial communications to a computer and logged into a text file for post-analysis.
In the laboratory experiments, we used a TSI 8530 DustTrak DRX aerosol
monitor with a PM
The station also included a nephelometer to monitor atmospheric visibility
in terms of the particle back-scatter (BSP) coefficient, reported in units
of Mm
The laboratory experiments were carried out in a 1 m
The field measurements were carried out at an air quality monitoring
station, situated close to a busy road, carrying approximately 100 vehicles
per min during the day. The PMS1003 was housed in a sealed weather-proof box
of dimensions
With the steady introduction of ambient air, the PM
The PM
As observed in the figure, the PM
Figure 3 shows the time series of the PM
Time series of the PM
Figure 4 shows the hourly PM
The PNC values reported by the PMS1003 in all size bins were also higher
during periods of fog. Under stable conditions, the PNCs reported by the
PMS1003 in the various size bins are generally linearly related. In Fig. 5,
we show the number concentration of particles larger than 1.0
The hourly PM
Next, we compare the PM
Graph of PNC > 1.0
Figure 7 shows the corresponding PNCs reported by the PMS1003 at 03:00, 06:00,
09:00 and 12:00 h on the day shown in Fig. 6b. The bars represent the
particle number dL
It is well known that humid air can have a negative effect on the performance of electronic circuits. For example, moisture in the air can decrease the insulation resistance in electrolytic capacitors and increase the leakage currents in transistors and integrated circuits, reducing the gain. In our previous tests (Jayaratne et al., 2018), we showed that the performance of some low-cost particle sensors such as the Sharp GP2Y and the Shinyei PPD42NS were affected at relative humidity as low as 50 %. The adverse effect was a fluctuation of the output signals, rather than a steady increase with humidity. This was obviously not due to particle growth, and we conclude that the electronics or optical characteristics were, in some way, responsible for these effects.
Variation of the PM
However, sensors such as the Plantower PMS1003, Innociple PSM305 and the
Nova SDS011, as well as particle monitors such as the TSI DustTrak, did not
show a marked effect until the relative humidity exceeded about 75 %, when
they began to show a steady increase. The results of the present study, with
the PMS1003 and the DustTrak showed that this was due to particle growth.
When the relative humidity is high, particle growth and fog are detected and
reported by particle monitoring instruments that do not have drying
facilities at the sample inlets. This effect needs to be taken into
consideration when using low-cost particle sensors, especially in
environments that contain hygroscopic salts such as near coastal regions.
Particles in the air begin to grow once the deliquescence relative humidity
is exceeded. For example, two hygroscopic salts that are commonly found in
Brisbane air are sodium chloride and ammonium sulfate. These have
deliquescence points of approximately 74 and 79 %, respectively (Hu et
al., 2010; Wise et al., 2007). Aerosol particles that contain these
substances will absorb moisture and grow when the relative humidity exceeds
these values. Our observations are in good agreement with these studies. The
high PM
PNCs reported by the PMS1003 in the six size bins at three hourly intervals during a morning with fog (30 July). Fog was observed between 03:00 and 06:30. The table under the figure gives additional information at the respective times.
An obvious question that arises from this work is whether it is possible to derive a correction factor for the particle number and mass concentrations reported by the low-cost sensors in the presence of high humidity and fog. Our results show that, once the deliquescence point is exceeded, the particle number and mass concentrations begin to increase and are not directly related to the absolute value of the relative humidity. Once the ambient temperature reaches the dew point temperature, the conditions become suitable for the formation of fog droplets in the air and, as a significant fraction of these water droplets fall within the detection size of the PMS1003 (Fig. 7), they are detected as particles. We also observed that the PNC and PM concentrations reported by the PMS1003 decreased in the presence of rain. This is not unexpected as it is known that rain washes out a fraction of airborne particles. More interestingly, our results show that the decrease in PNC and PM concentrations reported by the PMS1003 due to rain were significantly greater when there was an episode of fog than when there was no fog. While a significant number of fog droplets fall within the detection size range of the PMS1003, almost all the rain drops are larger than the maximum detection size of particles. We hypothesise that the raindrops were washing out the fog droplets in the air, resulting in an overall decrease in the reported PNC and PM concentrations reported by the low-cost particle sensors that have no drying facilities at their sample inlets. Moreover, the relative humidity of the atmosphere increased during rain, often approaching 100 %. Raindrops are too large to be detected by most particle sensors and, as such, they do not show an increase in concentration during rain. For these reasons, we find that there is no direct relationship between the relative humidity in the atmosphere and the PNC and PM concentrations reported by a sensor or monitor with no drying facility at its inlet and, as such, it is not possible to derive any appropriate correction factors for this effect.
As they generally do not have drying facilities at their sample inlets, low-cost particle sensors measure what is actually present in the air, including both the solid and liquid phases of the particles. This is a real observation and not an artefact of the instrument, as suggested by Crilley et al. (2018). This is an important aspect to be kept in mind when using low-cost sensors to assess the pollution levels in the atmosphere. What this illustrates is that it should not be presumed that low-cost sensors are suited for regulatory applications. For example, while it is reasonable to use low-cost sensors to measure the actual particle mass concentrations that are present in the air; such observations should not be used to verify if the air quality meets the stipulated guidelines or standards for particle pollution.
Data used in this paper may be obtained by contacting the Corresponding Author.
RJ Designed the project, carried out the measurements, analysed the data and wrote the paper. XL carried out the measurements and processed the data. PT assisted with the measurements, provided scientific input. MD designed and constructed the low cost PM sensor package. LM supervised the project and provided scientific input.
The authors declare that they have no conflict of interest.
We would like to thank the Queensland Department of the Environment and Science for providing the facilities and data from the air quality monitoring station. This study was supported by linkage grant LP160100051 from the Australian Research Council. We are grateful for useful discussions with Graham Johnson and Gavin Fisher. Akwasi Asumadu-Sakyi, Mawutorli Nyarku, and Riki Lamont assisted with the field work. Edited by: Pierre Herckes Reviewed by: two anonymous referees