Evaluation of Emerging Air Sensor Performance
On this page:
- What do the sensor terms used in these tables mean?
- Where can I get information about the air pollutants measured by these sensors?
- Related links
Emerging air quality sensors – with general traits of being more compact, directly reading pollutants, and lower in cost than traditional methods - have a wide appeal to professional researchers, community groups, students, and citizen scientists alike. Since this technology is still under development, little information exists on the quality of data that these sensors produce.
To help those interested in using sensors as part of air monitoring projects, EPA researchers evaluate sensors for how well they measure air pollutants and how easy they are to use. Placing the sensors near a regulation grade monitor, researchers collect data on air quality with both technologies. By assessing these data collected under the same air quality and weather conditions, researchers can compare how accurate and reliable low-cost technologies are to traditional methods.
The following air sensors have been evaluated in laboratory or field studies by EPA researchers. To assist the user, Sensor Evaluation Tables have been created that provide selected sensor performance results from reports, which can be found on the Air Sensor Toolbox. Note that the results provided in these tables are a subset of all of the sensors and tests conducted, some results are not shown in these summary tables for sensors that had inconclusive results due to either data logging failure, reference monitor failure, or other limiting factors. The full information on all test results are available in the full-length reports. EPA researchers continue to test sensors in various field or laboratory studies, and emerging results will be added to these tables over time.
Any mention of trade names or commercial products does not constitute endorsement. Note that cost information is not reported here, as the market prices of sensors are at the purview of the manufacturer or distributors, and may change with time or purchasing volume.
What do the sensor terms used in these tables mean?
- Detection approach – Explains how the sensor measures a particular pollutant.
- Operating details – Additional information on the general design of the device, data output rate, and data units
- Test results
- Testing Environment – Explains how the test was conducted. Field test environments include information on the pollution levels experienced during the test. Laboratory test environments provide the concentration ranges tested.
- Reference monitor used – Information on the federal reference or equivalent method used as a comparison point against the sensor.
- Averaging (Avg) time; testing period – Information on how the sensor and reference monitor data were averaged and how long the test was conducted.
- R2 – R2 is a statistical term. For these studies, it represents how well a sensor performs in comparison to the reference monitoring technologies. A value of R2=1 would mean the sensor has perfect correlation to the reference monitor, and R2=0 would mean no correlation to a reference monitor.
Where can I get information about the air pollutants measured by these sensors?
Particulate matter (PM) sensors
Sensor model |
Detection approach |
Operating details |
Test Results... |
||||
---|---|---|---|---|---|---|---|
Test Environment |
Reference monitor used |
Avg time; testing period |
R2 |
Citation |
|||
Alphasense OPC N2 |
Optical particle counting (0.38 to 17 microns). The cumulative particle counts across various size designations are converted to estimates of PM2.5 and PM10 particle mass concentrations. |
Unit was integrated into a prototype United Nations multi-pollutant sensor pod. Designed as a stationary monitor; it recorded data as 1 minute averages in units of µg/m3. |
One month (November 2016) of continuous testing at an EPA testing platform in Research Triangle Park, NC |
GRIMM EDM 180 FEM PM2.5 Monitor |
1 h averaging period; Approximately 1 month of continuous data collections |
0.007 (PM2.5) 0.01 (PM10) |
|
Shinyei | Volume scattering – particles (possibly including particles larger than 2.5 µm) entering the sensor scatter light from an internal light source. The scattered light signal is converted to an estimated particle mass concentration. | Designed as a stationary miniaturized monitor; can record data as fast as 1 second; data units in µg/m3. | Minimum 30 day testing period of duplicate or triplicate monitors at a state regulatory monitoring site in hot and humid conditions in Decatur, GA. | MetOne BAM 1020 FEM PM2.5 Monitor | 12 h averaging period; minimum 30 days | 0.45 to 0.60 | Community Air Sensor Network (CAIRSENSE) project |
Dylos | Optical particle counter – particles entering sensor are individually sized and counted based on how they scatter light. The sensor outputs particle counts in two size ranges (>0.5 µm; >2.5 µm). | Designed for indoor use; can record data as fast as 1 minute; data output units in particle counts. | Minimum 30 day testing period of duplicate or triplicate monitors at a state regulatory monitoring site in hot and humid conditions in Decatur, GA. | MetOne BAM 1020 FEM PM2.5 Monitor | 12 h averaging period; minimum 30 days | 0.63 to 0.67 (pro); 0.58 (DC1100) | Community Air Sensor Network (CAIRSENSE) project |
AirBeam | Volume scattering – particles (possibly including particles larger than 2.5 µm) entering the sensor scatter light from an internal light source. The scattered light signal is converted to an estimated particle mass concentration. | Designed as a highly portable handheld monitor; data are reported in units of µg/m3. | Minimum 30 day testing period of duplicate or triplicate monitors at a state regulatory monitoring site in hot and humid conditions in Decatur, GA. | MetOne BAM 1020 FEM PM2.5 Monitor | 12 h averaging period; minimum 30 days | 0.65 to 0.66 | Community Air Sensor Network (CAIRSENSE) project |
MetOne | Optical particle counter – particles entering sensor are individually sized and counted based on how they scatter light. The sensor outputs estimated mass concentrations in four size fractions (PM1, PM2.5, PM4, and PM10). | Designed as a handheld monitor; can record data as fast as 1 minute; data output units in micrograms per cubic meter (µg/m3). | Minimum 30 day testing period of duplicate or triplicate monitors at a state regulatory monitoring site in hot and humid conditions in Decatur, GA. | MetOne BAM 1020 FEM PM2.5 Monitor | 12 h averaging period; minimum 30 days | 0.32 to 0.41 | Community Air Sensor Network (CAIRSENSE) project |
Air Quality Egg | Volume scattering – particles (possibly including particles larger than 2.5 µm) entering the sensor scatter light from an internal light source. The scattered light signal is converted to an estimated particle mass concentration. | Designed for indoor use or outdoor use with proper weather shielding. Data are reported in units of µg/m3. | Minimum 30 day testing period of duplicate or triplicate monitors at a state regulatory monitoring site in hot and humid conditions in Decatur, GA. | MetOne BAM 1020 FEM PM2.5 Monitor | 12 h averaging period; minimum 30 days | -0.06 to 0.40 | Community Air Sensor Network (CAIRSENSE) project |
Cairpol CairClip PM - prototype |
Volume scattering – particles (possibly including particles larger than 2.5 microns (µm) entering the sensor scatter light from an internal light source. The scattered light signal is converted to an estimated particle count concentration. |
Lightweight and miniature; can record data as fast as 1 minute; data output in µg/m3. |
Wintertime outdoors in Durham, North Carolina; Reference monitor PM2.5 ranged ~2-45 µg/m3. |
Grimm Model EDM180 PM2.5 monitor |
5 min; ~1.5 months |
0.06 |
|
Airviz Speck v2 |
Volume scattering – particles (possibly including particles larger than 2.5 µm) entering the sensor scatter light from an internal light source. The scattered light signal is converted to an estimated particle count concentration. |
Designed for indoor use; can record data as fast as 5 seconds; data output in particle counts. |
Wintertime outdoors in Durham, North Carolina; Reference monitor PM2.5 ranged ~2-23 µg/m3. |
Grimm Model EDM180 PM2.5 monitor |
5 min; ~1.5 months |
0.01 |
|
Dylos DC1100 |
Optical particle counter – particles entering sensor are individually sized and counted based on how they scatter light. The sensor outputs particle counts in two size ranges (>0.5 µm; >2.5 µm). |
Designed for indoor use; can record data as fast as 1 minute; data output units in particle counts. |
Wintertime outdoors in Durham, North Carolina; Reference monitor PM2.5 ranged ~2-45 µg/m3. |
Grimm Model EDM180 PM2.5 monitor |
5 min; ~1.5 months |
0.55a |
|
Met One Model 831 |
Optical particle counter – particles entering sensor are individually sized and counted based on how they scatter light. The sensor outputs estimated mass concentrations in four size fractions (PM1, PM2.5, PM4, and PM10). |
Designed as a handheld monitor; can record data as fast as 1 minute; data output units in micrograms per cubic meter ( µg/m3). |
Wintertime outdoors in Durham, North Carolina; Reference monitor PM2.5 ranged ~2-45 µg/m3. |
Grimm Model EDM180 PM2.5 monitor |
5 min; ~1.5 months |
0.77b |
|
RTI MicroPEM |
Particles enter through size-selective inlet that removes particles >2.5 µm, then the remaining particles scatter light from a light source. An integrated filter collects all the particles, which can be optionally weighed in a laboratory after a period of use. |
Designed as a wearable monitor for indoor or outdoor environments; can record data as fast as 10 seconds; data units in µg/m3. |
Summertime outdoors in Durham, North Carolina; Reference monitor PM2.5 ranged ~4-33 µg/m3. |
Grimm Model EDM180 PM2.5 monitor |
5 min; ~1 month |
0.72 |
|
Shinyei PMS-SYS-1 |
Volume scattering – particles (possibly including particles larger than 2.5 µm) entering the sensor scatter light from an internal light source. The scattered light signal is converted to an estimated particle count concentration. |
Designed as a stationary miniaturized monitor; can record data as fast as 1 second; data units in µg/m3. |
Fall outdoors in Durham, North Carolina; Reference monitor PM2.5 ranged ~2-26 µg/m3. |
Grimm Model EDM180 PM2.5 monitor |
5 min; ~1 month |
0.15 |
|
Perkin-Elmer Elm |
Volume scattering – particles (possibly including particles larger than 2.5 µm) entering the sensor scatter light from an internal light source. The scattered light signal is converted to an estimated particle mass concentration. |
Designed as a stationary outdoor monitor; can record data as fast as 1 minute; data units in µg/m3. |
Wintertime outdoors in Durham, North Carolina; Reference monitor PM2.5 ranged ~2-23 µg/m3. |
Grimm Model EDM180 PM2.5 monitor |
5 min; ~1.5 months |
0.00 |
a Comparison results are for the small channel (>0.5 um).
b Comparison results shown are for the PM1 channel, which had the highest correlation with the reference monitor. The PM2.5 channel had significant outliers that were unexplained.
Gas Phase Sensors
Sensor model (pollutant types) |
Detection approach |
Operating details |
Test Results |
||||
---|---|---|---|---|---|---|---|
Test Environment |
Reference monitor used |
Avg time; testing period |
R2 |
Citation |
|||
AQMesh (ozone) | All gases are detected by air passing over electrochemical cells. | Measures NO, NO2, CO, CO2, SO2, O3 (all in ppb). | Minimum 30 day testing period of duplicate or triplicate monitors at a state regulatory monitoring site in hot and humid conditions in Decatur, GA. | Thermo Fisher Scientific FEM 49I ozone monitor | hourly comparisons, minimum 30 days | 0.39 to 0.45 | Community Air Sensor Network (CAIRSENSE) project |
CairClip (ozone) | The sensor pulls in a controlled air flow, which passes over a single electrochemical cell that responds to both nitrogen dioxide (NO2) and O3. To isolate one pollutant or the other, a separate measurement is needed in field tests. In laboratory tests, one gas type is introduced to test the sensor response. | The sensor device is miniature and designed for either portable use or stationary use with purchase of accessory peripherals; can record data as fast as 1 second; data units in ppb. | Minimum 30 day testing period of duplicate or triplicate monitors at a state regulatory monitoring site in hot and humid conditions in Decatur, GA. | Thermo Fisher Scientific FEM 49I ozone monitor | hourly comparisons, minimum 30 days | 0.82 to 0.94 | Community Air Sensor Network (CAIRSENSE) project |
Aeroqual SM50 (ozone) | Gas-sensitive semiconductor (GSS). | Measures O3 in ppm. | Minimum 30 day testing period of duplicate or triplicate monitors at a state regulatory monitoring site in hot and humid conditions in Decatur, GA. | Thermo Fisher Scientific FEM 49I ozone monitor | hourly comparisons, minimum 30 days | 0.91 to 0.97 | Community Air Sensor Network (CAIRSENSE) project |
CairClip (nitrogen dioxide) | The sensor pulls in a controlled air flow, which passes over a single electrochemical cell that responds to both nitrogen dioxide (NO2) and O3. To isolate one pollutant or the other, a separate measurement is needed in field tests. In laboratory tests, one gas type is introduced to test the sensor response. | The sensor device is miniature and designed for either portable use or stationary use with purchase of accessory peripherals; can record data as fast as 1 second; data units in ppb. | Minimum 30 day testing period of duplicate or triplicate monitors at a state regulatory monitoring site in hot and humid conditions in Decatur, GA. | Thermo Fisher Scientific FEM 42C nitrogen dioxide monitor | hourly comparisons, minimum 30 days | 0.42 to 0.76 | Community Air Sensor Network (CAIRSENSE) project |
AQMesh (nitrogen dioxide) | All gases are detected by air passing over electrochemical cells. | Measures NO, NO2, CO, CO2, SO2, O3 (all in ppb). | Minimum 30 day testing period of duplicate or triplicate monitors at a state regulatory monitoring site in hot and humid conditions in Decatur, GA. | Thermo Fisher Scientific FEM 42C nitrogen dioxide monitor | hourly comparisons, minimum 30 days | 0.14 to 0.32 | Community Air Sensor Network (CAIRSENSE) project |
Air Quality Egg (nitrogen dioxide) | The sensor device passes air over an internal metal oxide sensor, which provides a resistance change that is converted to an estimated NO2 concentration. | Designed for indoor use or outdoor use with proper weather shielding. Data are reported in units of ppb. | Minimum 30 day testing period of duplicate or triplicate monitors at a state regulatory monitoring site in hot and humid conditions in Decatur, GA. | Thermo Fisher Scientific FEM 42C nitrogen dioxide monitor | hourly comparisons, minimum 30 days | -0.25 to -0.22 | Community Air Sensor Network (CAIRSENSE) project |
WT-SU1 Dynamo (ozone) |
Ozone (O3) is detected via a metal oxide semiconductor. |
Sensor is incorporated into a weather station; data are output in units of parts per billion (ppb). |
Laboratory test in a glass exposure chamber with a gas standard introduced at controlled concentrations; Reference monitor O3 ranged 0-400 ppb. |
2B Model 205 O3 Analyzer |
1 min; several hours |
O3: 0.95 |
|
CairClip NO2/O3, USB version (ozone) |
The sensor pulls in a controlled air flow, which passes over a single electrochemical cell that responds to both nitrogen dioxide (NO2) and O3. To isolate one pollutant or the other, a separate measurement is needed in field tests. In laboratory tests, one gas type is introduced to test the sensor response. |
The sensor device is miniature and designed for either portable use or stationary use with purchase of accessory peripherals; can record data as fast as 1 second; data units in ppb. |
Laboratory test in a glass exposure chamber with a gas standard introduced at controlled concentrations; Reference monitor O3 ranged 0-400 ppb; NO2 ranged 0-200 ppb. |
2B Model 205 O3 Analyzer; Thermo Model 42C NO/NO2/NOx Analyzer |
1 min; several hours |
O3: 1.0 NO2: 1.0 |
|
AirCasting (nitrogen dioxide) |
The sensor device passes air over an internal metal oxide (MiCS-271) sensor, which provides a resistance change that is converted to an estimated NO2 concentration. |
The sensor device is miniature and designed for portable use, integrated with a smartphone app; can record data as fast as 1 minute; data units in ppb. |
Laboratory test in a glass exposure chamber with a gas standard introduced at controlled concentrations; Reference monitor NO2 ranged 0-200 ppb. |
Thermo Model 42C NO/NO2/NOx Analyzer |
1 min; several hours |
NO2: 0.98 |
|
Platypus (nitrogen dioxide) |
Nitrogen dioxide is detected using a thin film liquid crystal mounted to a metal strip. Each measurement of NO2 requires a new material strip to be entered into the device. |
Designed as a handheld monitor; each data point requires a manual filter change, which can be conducted at variable time intervals; sensor output is in units of ppb. |
Laboratory test in a glass exposure chamber with a gas standard introduced at controlled concentrations; Reference monitor NO2 ranged 0-200 ppb. |
Thermo Model 42C NO/NO2/NOx Analyzer |
Several minutes of sample time per test interval (5 concentrations tested) |
NO2: 0.80 |
|
CitiSense (nitrogen dioxidea) |
All gases are detected by air passing over electrochemical cells (O3: O3-3E-1 from CityTechnology; CO: NO2: NO2-AQ from Alphasense), which provides a resistance change that is converted to an estimated concentration. |
Sensor device is designed for portable use; integrated with a smartphone app; can record data as fast as 6 seconds; data units in ppb. |
Laboratory test in a glass exposure chamber with a gas standard introduced at controlled concentrations; Reference monitor O3 ranged 0-400 ppb; NO2 ranged 0-200 ppb. |
Thermo Model 42C NO/NO2/NOx Analyzer |
1 min; several hours |
NO2: 0.98 |
|
U-Pod (nitrogen dixode, ozoneb) |
The U-Pod, at the time of testing, detects NO2 and O3 via a metal oxide sensors (NO2: MiCS-2710, O3: MiCS-2611 from SGX SensorTech), which provides a voltage output that is converted to concentration units. |
Sensor device is designed for stationary use; can record data as fast as 5 seconds; data units in ppb. |
Laboratory test in a glass exposure chamber with a gas standard introduced at controlled concentrations; Reference monitor O3 ranged 0-400 ppb; NO2 ranged 0-200 ppb. |
2B Model 205 O3 Analyzer; Thermo Model 42C NO/NO2/NOx Analyzer |
1 min; several hours |
O3: 0.88 NO2: n/a |
|
Unitec SENS-IT (benzene) |
The SENS-IT passes air flow over a thick film metal oxide semiconductor, which generates a signal that is converted to a specific gas concentration. |
Sensor is miniature, designed for incorporation into stationary or mobile monitoring applications; data at the time of testing were output in voltage units. |
Laboratory test in a stainless steel test chamber with specific gases (benzene, 1,3 butadiene, tetrachloroethylene) introduced individually or in mixtures; concentration ranged 0-25 ppb. |
Gas chromatography |
~3 hours; 18 hours |
Benzene-only: 0.90 |
Next Generation Air Monitoring (NGAM) VOC Sensor Evaluation Report |
a The CitiSense monitor also includes sensors for carbon monoxide and ozone, which were not evaluated in the laboratory test.
b The U-Pod also includes sensors for nitric oxide, carbon monoxide, volatile organic compounds, carbon dioxide, and sulfur dioxide, which were not evaluated in this laboratory test.
CAIRSENSE-Denver Study
The Community Air Sensor Network (CAIRSENSE) project was a multi-year research project that focused on evaluating performance characteristics and limitations of low-costs sensors. For the second part of the CAIRSENSE study, Denver, Colorado, was chosen to test the sensors’ performance under conditions of high altitude, dry and lower temperature conditions. Beyond assessing sensor performance through correlation with a reference monitor, this study also investigated the degree to which data from sensors is able to produce similar temporal, wind-direction, and transient event trends in comparison to a high time-resolution reference monitor. Citation: Atmos. Meas. Tech. Discuss., https://doi.org/10.5194/amt-11-4605-2018
CAIRSENSE-Denver Sensor Data Completeness
Sensor | Measurement Completeness | Sensor Logging Error | Percent Missing Data |
---|---|---|---|
Aeroqual | 82% | 0% | 18% |
73% | 0% | 27% | |
81% | 5% | 13% | |
Air Assure | 87% | 0% | 13% |
87% | 0% | 13% | |
87% | 0% | 13% | |
AirBeam | 74% | 0% | 25% |
62% | 6% | 32% | |
62% | 6% | 32% | |
CairClip | 29% | 53% | 18% |
63% | 13% | 24% | |
63% | 23% | 13% | |
Dylos | 82% | 0% | 18% |
82% | 0% | 18% | |
72% | 1% | 27% | |
OPC-N2 | 77% | 0% | 23% |
76% | 0% | 24% | |
71% | 0% | 29% | |
Shinyei | 82% | 0% | 18% |
73% | 0% | 27% | |
87% | 0% | 13% | |
Speck | 92% | 0% | 8% |
93% | 0% | 7% | |
96% | 0% | 4% | |
TZOA | 61% | 0% | 39% |
47% | 0% | 53% | |
47% | 0% | 53% |
Regression and Precision Results for CAIRSENSE sensors (1-hour time averaged)
Sensor | Pollutant | Average Reference Concentration1 | Regression Slope | Regression Intercept | Pearson Correlation, R | RMS (Root Mean Square) | Number of Hourly Measurements | |
---|---|---|---|---|---|---|---|---|
Precision | ||||||||
(%) | ||||||||
Aeroqual SM-50 | O3, ppb | 18.8 ppb | 0.56 | -0.004 | 0.93 | 73 | 3325 | |
0.58 | -0.004 | 0.92 | 2963 | |||||
0.77 | -0.004 | 0.96 | 3279 | |||||
TSI Air Assure | PM, µg/m3 | 7.8 µg/m³ | 1.14 | 2.64 | 0.8 | 41 | 3486 | |
1.13 | -0.04 | 0.78 | 3486 | |||||
1.19 | -1.38 | 0.81 | 3486 | |||||
AirCasting AirBeam | Particle Count, hundreds of particles per cubic foot (hppcf) | 7.8 µg/m³ | 273 | -323 | 0.82 | 6 | 3028 | |
278 | -124 | 0.84 | 2539 | |||||
322 | -352 | 0.82 | 2532 | |||||
Cairpol CairClip | O3, ppb | 18.8 ppb | NA2 | NA2 | NA2 | NA2 | 738 | |
-0.04 | -23.6 | -0.06 | 2831 | |||||
1.03 | -39 | 0.46 | 2852 | |||||
Cairpol CairClip | NO2, ppb | 26.8 ppb | NA2 | NA2 | NA2 | NA2 | 738 | |
0.65 | -10 | 0.87 | 2831 | |||||
0.67 | -15 | 0.84 | 2852 | |||||
Dylos DC1100/DC1100 Pro | "Small" Particle Count, hppcf | 7.8 µg/m³ | 64 | -152 | 0.86 | 15 | 3324 | |
428 | -1182 | 0.78 | 3324 | |||||
431 | -941 | 0.73 | 2937 | |||||
Dylos DC1100/DC1100 Pro | "Large" Particle Count, hppcf | 12.0 µg/m³ | 1.3 | 5.5 | 0.4 | 10 | 3324 | |
5.7 | 73 | 0.33 | 3324 | |||||
4.9 | 84 | 0.27 | 2937 | |||||
Alphasense OPC-N2 | PM2.5, µg/m3 | 7.8 µg/m³ | 0.4 | -0.3 | 0.45 | 108 | 2969 | |
0.49 | -1.66 | 0.34 | 2939 | |||||
0.07 | 0.6 | 0.11 | 2735 | |||||
Alphasense OPC-N2 | PM10, µg/m3 | 19.6 µg/m³ | 0.45 | 2.98 | 0.47 | 101 | 2969 | |
0.54 | -1.06 | 0.68 | 2939 | |||||
0.12 | 2.86 | 0.2 | 2735 | |||||
Shinyei PMS-SYS-1 | PM2.5, µg/m3 | 7.8 µg/m³ | 0.58 | 0.24 | 0.71 | 20 | 3325 | |
0.54 | 0.8 | 0.72 | 2963 | |||||
0.42 | 4.35 | 0.01a | 3486 | |||||
Airviz Speck | PM2.5, µg/m3 | 7.8 µg/m³ | 0.76 | 13 | 0.24 | 37 | 3557 | |
0.74 | 15 | 0.4 | 3584 | |||||
0.62 | 10 | 0.35 | 3971 | |||||
TZOA PM Research Sensor | Particle Count, hppcf | 7.8 µg/m³ | NA2 | NA2 | NA2 | 17b | 2341 | |
6.68 | 1.37 | 0.66 | 1838 | |||||
6.75 | 2.16 | 0.72 | 1836 |
1Average Concentration calculated for hours with valid sampling data.
2Correlation results not shown due to large amount of missing or invalid data
a Shinyei Unit 3’s correlation improved to 0.84 when only considering data from October 16 and later
bTZOA Unit 1 was excluded from RMS precision calculations