In the field of air quality measurement, low-cost sensors have been extensively used for capturing high-resolution spatial and temporal pollutant concentrations. However, data accuracy and the impact of environmental conditions particularly temperature and relative humidity, remain a significant issue. Careful sensor calibrations against reference instruments are necessary for ensuring data accuracy and validity. This study investigated the effectiveness of multiple calibration methodologies (linear calibration, 3rd order polynomial calibration, and Random Forest, RF) for low-cost air quality sensor calibration across six cities (Atlanta, Portland, Riverside, Sacramento, New York City, and Phoenix) in the United States. The data for this study is collected from April 2019 to April 2020 in the corresponding cities. The study showed that the local calibration provides better accuracy than adopting calibration equation developed for other regions across all the models tested. The potential reasons could be that different cities may have different pollution source combinations, which may result in varying types and concentration levels of interfering gas. Different weather conditions will also contribute to such results, as parameters especially temperature and relative humidity are known to impact the performance of low-cost sensors. Additionally, if included CO signal in the calibration models of NO2 and O3, model performances are found to increases across all the cities. Low-cost CO sensor is known to general perform better and has better long-term stability than their NO2 and O3 sensor counterparts. Results of this thesis contribute to better understanding on low-cost air pollutant sensor deployment and calibration techniques.
| Identifer | oai:union.ndltd.org:ucf.edu/oai:stars.library.ucf.edu:etd2020-2818 |
| Date | 15 August 2023 |
| Creators | Hasan, Md Hasibul |
| Publisher | STARS |
| Source Sets | University of Central Florida |
| Language | English |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | Electronic Theses and Dissertations, 2020- |
Page generated in 0.0156 seconds