Analyzing Algorithm Accuracy and Bay Breeze Influences on Purple Air Sensors in Baltimore

Purple Air sensors are low-cost air quality sensors used by community scientists which have led to a significant increase in air quality observations across the world. However, Purple Air data has been shown to overestimate PM 2.5 concentrations when compared to the EPA’s Air Quality Monitoring Network. Previous research has shown that Purple Air’s performance is strongly dependent on environmental factors such as humidity and abundance of coarse particles which vary seasonally. This research seeks to identify the specific environmental conditions contributing to measurement errors in Purple Air sensors, focusing on the impact of humidity, temperature and dewpoint temperature in the Chesapeake Bay area, particularly in Baltimore, Maryland. Using data collected from a 2.5 year period, a nationwide correction model based on Clements et al 2020 was applied to the Purple Air sensor data. Purple Air sensors have different algorithms in processing the PM 2.5 concentration values detected. This research specifically focuses on the CF_1 and ALT_CF3 algorithm and compares the results. The accuracy of Purple Air sensors in measuring PM 2.5 concentration is heavily influenced by the algorithm used by the Plantower PMS5003 sensor. The goal of this research is to better understand the underlying cause of PM 2.5 measurement errors in Purple Air sensors, which can inform future improvements in calibration and data interpretation. By focusing on the relationship between measurement errors and meteorological factors, this research contributes to a more nuanced understanding of air quality data accuracy in Baltimore. The findings confirm previous research claims that the CF_1 algorithm significantly overestimates PM 2.5 concentrations while the ALT_CF3 is less susceptible to environmental biases. The nationwide correction factor is more effective when applied to the ALT_CF3 data, especially in reducing the bias of relative humidity on Purple Air’s PM 2.5 overestimation. Continued refinement of these algorithms, along with better sensor collocations and regional correction factors is necessary to ensure more reliable and consistent data from Purple Air sensors across diverse settings.