Calibration of indoor temperature and relative humidity readings in the PurpleAir monitor

Temperature extremes are associated with a variety of negative health outcomes, but since people living in developed areas of the world spend most of their time indoors, outdoor temperatures are a poor substitute for personal exposure assessment. And the importance of accurate indoor temperature measurement has only become more apparent alongside the growing impact of climate change on the frequency and intensity of temperature extremes on public health. The development and implementation of low-cost sensors have improved economic and practical feasibility of in-home exposure assessment for temperature and a variety of indoor contaminants. One example is the PurpleAir particulate matter sensor. However, onboard electronics likely bias measured temperature (T) and relative humidity (RH) values in such devices. The objectives of this investigation were to (1) characterize bias and error of T and RH estimates by comparing 30,936 hourly mean values to a dedicated, calibrated temperature and RH sensor co-located at 115 homes in Baltimore, MD, (2) develop calibration equations to accurately approximate the reference values, (3) validate the performance of these equations, and (4) validate the transportability of these calibration equations using 7697 pairs of hourly mean temperature and RH values from 22 homes in western Maryland. The PurpleAir sensors measured higher temperatures and lower RH than indoor reference measurements. Calibration schemes using a bias correction and multiple linear regression were considered. Calibration of PurpleAir temperature in degrees C (TPA) may be performed by Tindoor = TPA - 3.95 degrees C (R2 = 0.80), and calibration of PurpleAir RH measured from 0-100% (RHPA) may be performed by RHindoor = - 0.29 + 1.39 center dot RHPA (R2 = 0.92). Calibration-corrected temperature and RH demonstrated low bias and error values, and these results were consistent in the withheld datasets.