Impact Analysis of Temperature and Humidity Conditions on Electrochemical Sensor Response in Ambient Air Quality Monitoring

Peng Wei, Zhi Ning, Sheng Ye, Li Sun, Fenhuan Yang, Ka-Chun Wong, Dane Westerdahl, Peter K. K. LouieSensors 2018 · OpenReview

TL;DR

This paper investigates the impact of temperature and humidity on low-cost electrochemical air quality sensors, demonstrating that while these sensors offer promise, their raw data requires significant correction to be reliable for ambient monitoring. The authors developed and validated several correction models, showing that accounting for environmental factors dramatically improves data accuracy, making these sensors more viable for research and citizen science applications.

What Problem Does This Paper Address?

The proliferation of low-cost air quality sensors presents an exciting opportunity for widespread, high-resolution air pollution monitoring. However, a major hurdle is the questionable data quality from these devices. Unlike expensive, laboratory-grade instruments that operate under controlled conditions, low-cost sensors are often deployed in real-world environments where factors like temperature and humidity can significantly influence their readings. This variability makes it difficult to trust the data for scientific research, regulatory purposes, or even for informed citizen science initiatives. This paper aims to bridge this gap by systematically evaluating the impact of temperature and humidity on common electrochemical sensors and developing robust correction methods to improve their accuracy in ambient conditions.

Methods

The study employed a multi-pronged approach:

  • Laboratory Calibration and Characterization: Four types of Alphasense B4 series electrochemical sensors (CO, NO, NO2, and Ox) were subjected to controlled laboratory tests. This involved exposing them to known concentrations of target gases under varying temperature (15-34 °C) and humidity (17-48%) conditions. The linearity of sensor response, cross-interference with other gases, and long-term drift (over two months) were assessed.
  • Correction Model Development: Based on the laboratory findings, three distinct correction models were developed. These models aimed to compensate for the observed environmental influences.
  • Model 0: A baseline model assuming a direct linear relationship between sensor output and gas concentration, without environmental correction.
  • Model 1: Incorporated temperature and humidity as independent variables affecting the sensor's sensitivity and baseline response.
  • Model 2 & 3: Explored different ways of utilizing both the active and reference electrode voltages, along with temperature and humidity, to refine the correction. Model 3, in particular, treated VRef as an independent parameter for T and RH correction.
  • Field Validation: The developed models were tested in a real-world urban environment in Hong Kong. Three sensor platforms were deployed alongside regulatory-grade reference monitors for an 8-day period. The performance of the uncorrected (Model 0) and the best-performing corrected models was compared against the reference data.
  • Error Analysis: Quantitative error analysis was performed using statistical metrics (R², standard deviation, mean error) and visual representations (histograms, scatter plots) to assess the effectiveness of the correction models in reducing deviations from reference measurements.

Datasets

  • Laboratory Data: Generated by exposing Alphasense B4 sensors to known concentrations of CO, NO, NO2, and O3 under controlled temperature and humidity conditions.
  • Field Data: Collected from deployed sensor platforms and co-located regulatory-grade air quality monitoring station (AQMS) data in Hong Kong over an 8-day period. This dataset captured real-world ambient conditions.

Key Insights and Ideas

  • Environmental Sensitivity is Crucial: Electrochemical sensors, even those with good linearity in controlled settings, are highly sensitive to temperature and humidity. This sensitivity is not uniform across all gases or sensor types.
  • Raw Data is Unreliable: Without correction, the data from these low-cost sensors can exhibit significant deviations from reference measurements, especially during periods of fluctuating environmental conditions.
  • Correction Models Make a Difference: Implementing T and RH correction models significantly improves the accuracy and reliability of sensor data. The choice of model depends on the specific sensor's response characteristics to environmental factors.
  • NO2/O3 Interaction: The study highlighted a specific challenge with the Ox sensor, which showed a strong response to NO2. This necessitates careful consideration of gas interactions and the use of filters or specific calculation methods (like subtracting NO2 from Ox readings) to accurately determine O3 concentrations.
  • Long-Term Drift is a Concern: While the study focused on short-term environmental impacts, it also observed sensor drift over a two-month period, underscoring the need for periodic recalibration.

Key Results

  • High Linearity in Lab: Sensors demonstrated excellent linearity (R² > 0.99) with target gases under stable laboratory conditions.
  • Variable Environmental Impact: The impact of temperature and humidity varied. For instance, the NO sensor's reference voltage showed a strong linear correlation with T and RH, while other sensors had less pronounced relationships.
  • Model Performance:
    • Model 1 (incorporating T and RH into sensitivity and baseline) showed improved performance for CO and NO2.
  • Model 3 (treating VRef as an independent parameter for T and RH correction) performed well for NO and O3.
  • Error Reduction:
    • The optimal corrected models reduced errors by up to 41% for CO, 35% for NO, 22% for NO2, and 32% for O3 compared to the uncorrected model.
  • Field validation showed that corrected models significantly reduced deviations from AQMS data, particularly during periods of large environmental variation. For example, corrected models improved sensor accuracy by 69% and 33% in specific T/RH bins for CO.

Limitations and Future Work

  • Short Field Test Duration: The field evaluation was limited to 8 days, which is insufficient for a comprehensive assessment of long-term drift.
  • Limited Humidity Range: The laboratory humidity tests were constrained to a maximum of 48%, and the system was unable to produce higher levels.
  • O3 Calculation Complexity: The calculation of O3 relied on subtracting NO2 from Ox readings, which can propagate errors if the NO2 measurement is inaccurate.
  • Future Work: The authors suggest further investigation into:
  • Long-term drift over periods of up to a year.
  • Applicability in extreme ambient conditions.
  • Developing more sophisticated methods to reduce the impact of long-term drift.
  • Generating more reliable data from these sensors, potentially through vendor collaboration and standardized protocols.
  • Exploring the performance of multi-gas monitoring systems in challenging environments.

Summary updated

Abstract

The increasing applications of low-cost air sensors promises more convenient and cost-effective systems for air monitoring in many places and under many conditions. However, the data quality from such systems has not been fully characterized and may not meet user expectations in research and regulatory uses, or for use in citizen science. In our study, electrochemical sensors (Alphasense B4 series) for carbon monoxide (CO), nitric oxide (NO), nitrogen dioxide (NO2), and oxidants (Ox) were evaluated under controlled laboratory conditions to identify the influencing factors and quantify their relation with sensor outputs. Based on the laboratory tests, we developed different correction methods to compensate for the impact of ambient conditions. Further, the sensors were assembled into a monitoring system and tested in ambient conditions in Hong Kong side-by-side with regulatory reference monitors, and data from these tests were used to evaluate the performance of the models, to refine them, and validate their applicability in variable ambient conditions in the field. The more comprehensive correction models demonstrated enhanced performance when compared with uncorrected data. One over-arching observation of this study is that the low-cost sensors may promise excellent sensitivity and performance, but it is essential for users to understand and account for several key factors that may strongly affect the nature of sensor data. In this paper, we also evaluated factors of multi-month stability, temperature, and humidity, and considered the interaction of oxidant gases NO2 and ozone on a newly introduced oxidant sensor.
BibTeX
@inproceedings{wei2018impact,
  author = {Peng Wei and Zhi Ning and Sheng Ye and Li Sun and Fenhuan Yang and Ka-Chun Wong and Dane Westerdahl and Peter K. K. Louie},
  title = {{Impact Analysis of Temperature and Humidity Conditions on Electrochemical Sensor Response in Ambient Air Quality Monitoring}},
  booktitle = {Sensors},
  year = {2018},
  url = {https://openreview.net/forum?id=TypKAtijO0},
}
Figure