Air Quality Monitor Calibration: A Step-By-Step Guide

Calibration is a central issue in the process of setting up and using DIY air quality sensors. It is the process of adjusting a sensor to a standard reference monitor or measurement to ensure the data it generates are roughly comparable to official data. However, this process is riddled with complexities and differing approaches, some of which are fervently fought over in communities that develop and use DIY technologies. Calibration can take place through hardware, hardware and software, or software and platforms, and it can involve testing sensors in situ, where environments and types of pollutants can inform the calibration process. The need for calibration defies any promise of universalism in networked sensing devices and demands that we challenge the universalisms of digital media as the sensor is always in process with its environment.

Calibration of low-cost PM2.5 monitor using a random forest model

Particle air pollution has been proven to have adverse health effects on humans. Low-cost monitoring among a large population group is an effective method for performing environmental health studies. However, concerns about the accuracy of low-cost monitors have hindered their widespread use in monitoring projects. This study aims to calibrate a low-cost particle monitor (HK-B3, Hike, China) through a controlled exposure experiment and establish an accurate calibration model to improve the applicability of low-cost air pollution monitors.

Methods

Our study used a MicroPEM monitor (RTI, America) as a standard particle concentration measurement device to calibrate the Hike monitors. We established a machine learning model to calibrate the particle concentration obtained by the low-cost PM2.5 monitors and used ten-fold validation to test the model. Additionally, we used a linear regression model to compare the results of the machine learning model.

Results

The values of the random forest model calibration results and observations were more condensed around the regression line y = 0.99x + 0.05, and the R-squared value (R^2 = 0.98) was higher than that for the linear regression (R^2 = 0.87). The random forest model showed better performance than the traditional linear regression model.

Our study provided an effective calibration method to support the accuracy of low-cost monitors. The machine learning method based on the calibration model can increase the effectiveness of future air pollution and health studies. By addressing the accuracy and consistency limitations of low-cost sensors, this model can enhance their implementation among large population groups.

Calibration of low-cost air quality sensors

The calibration process should occur before, during, and after data collection. This involves comparing the response of the low-cost sensor to a reference instrument or system. While calibration can significantly improve the accuracy of low-cost sensors, it is important to note that their precision may still not match that of reference-grade measurements.

One study proposed a spatial calibration and mapping approach for low-cost PM2.5 sensors based on reference-grade measurements at regulatory stations. The spatial calibration model, which is a local linear transformation, outperformed the non-spatial calibration model by reducing inconsistencies and improving the R-square value to 0.94. This approach is especially useful in environments with high relative humidity, where calibration is critical due to the influence of humidity on sensor performance.

Another project aimed to develop repeatable protocols for calibrating and interpreting air quality measurements from low-cost sensors. This included working with multiple sensors to account for variability and understanding how calibration changes over time and under different conditions.

It is important to note that low-cost sensors may not be factory-calibrated, and their components may have high variability. Therefore, calibration methods should be tailored to the specific type of sensor and the reasons for monitoring air quality.

Calibration of a dust sensor using a fry pan bacon test

The following is a guide on how to calibrate a dust sensor using the "Fry Pan Bacon Test". This method was popularised by Chris Nafis, an employee of General Electric Research, who suffered from allergies and asthma and subsequently became interested in local indoor and outdoor air quality.

Materials

• A Shinyei Model PPD42NS Dust Sensor supplied by Seeed Studio
• Arduino microcontroller
• Fry pan
• Bacon

Steps

• Assemble the Shinyei dust sensor with the Arduino microcontroller.
• Using the Arduino instructions, load the relevant code onto the microcontroller.
• Fry bacon in the fry pan.
• Compare the readings from the dust sensor with the known standards for the particulates produced by frying bacon.
• Adjust the dust sensor's settings until the readings are in line with the known standards.

Tips

• This method of calibration is particularly useful for calibrating dust sensors to measure particulate matter in urban environments.
• The Shinyei sensor is designed to measure particulate matter levels and is responsive to particulates with a diameter of 1 µm or larger.
• Calibration may need to be repeated throughout the life of the sensor to account for drift and other external effects.
• This calibration method can be used in conjunction with other methods, such as co-location with regulatory monitors, to increase accuracy.

Further Considerations

• The process of calibration is an ongoing practice that is entangled with the process of generating measurements.
• Calibration is a co-constitutive technical and political practice that shapes how data is mobilised for political change.
• Different communities may have differing motivations and preferred calibration techniques, which can lead to disagreements and debates.
• Calibration is not a universal process and may need to be adjusted for different environments and use cases.

Calibration of a Dustbox particulate matter 2.5 monitor

The Dustbox is a low-cost air quality monitor designed by Citizen Sense to enable citizens to undertake air quality sensing in their neighbourhoods. It measures small particles between 1 to 2.5 micrometres (μm), which are designated as particulate matter 2.5 (PM2.5). These particles can include soot, dirt, dust, pollen, mould, and carbon with volatile organic compounds attached.

The Dustbox uses a Plantower PMS5003 sensor with an infrared laser to optically sense and measure airborne particulate matter. It also has a small fan that draws air into the sensor. Before using the Dustbox, it is important to set up the device by downloading the Electric Imp app to your smartphone and adding your WiFi information. This allows the Dustbox to send data to the Citizen Sense platform.

To calibrate the Dustbox, it is necessary to work with atmospheric scientists or experts in the field. In the case of the Dustbox, Citizen Sense collaborated with atmospheric scientists at King's College London to calibrate the devices. The Dustboxes were also tested and adjusted at Goldsmiths, University of London, to ensure consistency across the devices.

It is worth noting that low-cost sensors like the Dustbox may require additional calibration techniques to improve accuracy. For example, in a Public Lab Research note, Ufuoma discusses the need for developing "readable and repeatable protocols for gathering, organising, calibrating and interpreting air quality measurements from low-cost sensors."

Therefore, when calibrating a Dustbox particulate matter 2.5 monitor, it is recommended to follow the manufacturer's instructions and seek expertise from atmospheric scientists or professionals in air quality monitoring to ensure accurate and reliable results.

Calibration of a low-cost air quality sensor using an adaptive neuro-fuzzy inference system

Calibration of air quality sensors is essential to ensure accurate and reliable measurements of air pollution levels. This is particularly important for low-cost sensors, which often suffer from data quality issues due to interference from meteorological conditions and other gas pollutants.

One approach to calibrating low-cost air quality sensors is through the use of an Adaptive Neuro-Fuzzy Inference System (ANFIS). This technique combines neural networks and fuzzy logic principles, offering the advantages of both methods. The ANFIS model can learn and adapt to new data, making it a powerful tool for calibrating sensors in dynamic and complex environments.

The calibration process using ANFIS typically involves several steps:

• Data visualisation and input selection: Choose relevant input parameters based on domain knowledge.
• Data preparation: Clean, normalise, and randomise the data. Split the data into training, testing, and validation sets.
• Model building: Train the ANFIS model using the training data. Optimise the model by adjusting parameters and removing irrelevant inputs.
• Model evaluation: Compare the predicted outputs of the ANFIS model with actual measurements to evaluate its accuracy and performance.

The ANFIS model has been applied to calibrate a custom-made low-cost air quality sensor, DiracSense, which monitors gas pollutants such as ozone (O3), nitrogen dioxide (NO2), and carbon monoxide (CO). The performance of the ANFIS model was compared with other calibration techniques, including a simple equation and supervised learning algorithms. The results showed that the ANFIS model had the lowest error and the highest correlation coefficients with respect to the reference instrument, demonstrating its effectiveness as a calibration tool.

The ANFIS technique has also been used to forecast air pollution levels by predicting the concentrations of atmospheric pollutants. In one study, the ANFIS model was trained using historical data on pollutant concentrations, meteorological parameters, and other factors. The model was then used to predict future pollutant levels, achieving a mean absolute error of less than 15%.

In summary, the ANFIS method offers a powerful and flexible approach to calibrating low-cost air quality sensors. By combining neural networks and fuzzy logic, the ANFIS model can learn and adapt to new data, making it suitable for dynamic and complex environments. The calibration process typically involves data preparation, model building, and evaluation steps to optimise the model's performance. The ANFIS technique has been successfully applied in various studies, demonstrating its effectiveness in improving the accuracy and reliability of air quality sensors.

Frequently asked questions